This module exploits a heap buffer overflow within versions of Exim prior to
version 4.69. By sending a specially crafted message, an attacker can corrupt the
heap and execute arbitrary code with the privileges of the Exim daemon.
The root cause is that no check is made to ensure that the buffer is not full
prior to handling '%s' format specifiers within the 'string_vformat' function.
In order to trigger this issue, we get our message rejected by sending a message
that is too large. This will call into log_write to log rejection headers (which
is a default configuration setting). After filling the buffer, a long header
string is sent. In a successful attempt, it overwrites the ACL for the 'MAIL
FROM' command. By sending a second message, the string we sent will be evaluated
with 'expand_string' and arbitrary shell commands can be executed.
It is likely that this issue could also be exploited using other techniques such
as targeting in-band heap management structures, or perhaps even function pointers
stored in the heap. However, these techniques would likely be far more platform
specific, more complicated, and less reliable.
This bug was original found and reported in December 2008, but was not
properly handled as a security issue. Therefore, there was a 2 year lag time
between when the issue was fixed and when it was discovered being exploited
in the wild. At that point, the issue was assigned a CVE and began being
addressed by downstream vendors.
An additional vulnerability, CVE-2010-4345, was also used in the attack that
led to the discovery of danger of this bug. This bug allows a local user to
gain root privileges from the Exim user account. If the Perl interpreter is
found on the remote system, this module will automatically exploit the
secondary bug as well to get root.
- jduck <firstname.lastname@example.org>
- hdm <email@example.com>