o      ooooo  oooo oooo     oooo      o      oooo   oooo
    888       888  88    8888o   888      888      8888o  88
   8  88        888      88 888o8 88     8  88     88 888o88
  8oooo88      88 888    88  888  88    8oooo88    88   8888
o88o  o888o o88o  o888o o88o  8  o88o o88o  o888o o88o    88


AxMan is a web-based ActiveX fuzzing engine. The goal of AxMan is to
discover vulnerabilities in COM objects exposed through Internet
Explorer. Since AxMan is web-based, any security changes in the
browser will also affect the results of the fuzzing process. This
allows for a much more realistic test than other COM-based assessment
tools. AxMan is designed to be used with Internet Explorer 6 only.


The latest version is 1.0.0


AxMan was developed by H D Moore <x[at]hdm.io> and is
available under the MIT license. The axscan.cpp component (and 
associated axman.exe) is based on Shane Hird's axenum tool included
with axfuzz (http://axfuzz.sf.net/). The latest version of AxMan can
be obtained from https://www.hdm.io/tools/axman/


AxMan works by first enumerating all registered COM objects and their
associated typelib information, then using that information to test
each object's properties and methods. To enumerate the COM
information, copy bin\axman.exe to the target system, and execute it
as an administrative user:

C:\> axman.exe myoutput

The axman.exe executable will take anywhere from 10 minutes to 2
hours to complete, depending on the number of COM objects registered
on your system. While it runs, a number of new processes will be
spawned and random popup windows will open. You can safely close any
popup windows as they open, or just wait for the entire process to
complete. Once axman.exe completes, the resulting directory should
contain one javascript source file for every registered COM object.

At this point, you should reboot the system in order to kill the
spawned COM processes and free system resources.

The next step is to prepare a web server to host the AxMan user
interface. A web server running on the local system is the most
efficient way to do this, but you can use any server capable of
serving plain HTML and javascript files. Copy the html subdirectory of
the AxMan archive to the web root and then copy the output directory
created by axman.exe to a subdirectory named 'conf'. This should
result in a directory structure like the following:


On the target system, open Internet Explorer and browse the index.html
page of the AxMan user interface. If everything has been installed
successfully, a status message will be shown indicating how many COM
object definitions were loaded.

Click the start button to begin the fuzzing process.


AxMan will trigger hundreds of crashes on a typical system. Trying to
determine what crashed and during what test is a challenge for a
browser-based fuzzing engine. My personal approach to install WinDbg
(available for free from Microsoft, search for 'Debugging Tools for
Windows') and attach to the iexplore.exe process prior to starting the
fuzzing process. After attaching the debugger, I hit F5 to continue
the process and switch back to Internet Explorer. AxMan will report
the current CLSID and tested property or method in the status bar of
the browser. This status bar is updated in real-time during the
testing process. I combine this with a 'tail' process on the log files
of the hosting web server to quickly identify which component crashed
and add the property or method to the blacklist. Keep in mind that the
Internet Explorer UI will not update while the method fuzzing stage
runs, which means that switching applications or doing basically
anything on the testing system will prevent you from seeing which
method triggered a specific crash.


The blacklist.js script contains a list of all objects, properties,
and methods to skip during the fuzzing process. This doubles as a
list of discovered bugs. As you find new vulnerabilities, update
the blacklist to exclude them in future runs. Please see the examples
in blacklist.js for more information.

In order to buy Microsoft some time, I have not included my AxMan
blacklist file with the current release. This blacklist took about
20 hours in total to create.


AxMan uses a unique string as part of both the property and method
fuzzing operations. This allows you to run tools like filemon and
regmon to determine what system-level operations happen when a
specific property is set or method is called. The default magic string
is set to 'AXM4N', but this can be changed in axman.js. Some sample
findings include:

Setting the PrinterName property for 2337A8C-E11D-11D0-BE48-00C04FC30DF6
results in spoolsv.exe access the following registry key:

Calling the parseURL method of 079aa557-4a18-424a-8eee-e39f0a8d41b9
results in the entire system path being scanned for a file matching
the argument name.


Server-side integration and asynchronous method testing are the two
big features that will be added to future versions. The current
implementation is very slow for methods with three arguments and may
cause the "Do you want to stop this script?" prompt to appear in
certain cases. Progress reporting needs an overhaul, possibly using
XMLHTTP and some form of server-side integration.