Two weeks ago (we're running late, sorry!) j00ru and I had the pleasure to attend one of the largest, most technical and renowned conferences in existence - Black Hat 2013 in Las Vegas, USA. The event definitely stood up to our expectations - the city was purely awesome, the venue was at least as great, we saw many interesting and truly inspiring talks and a whole bunch of old friends, not to mention meeting a fair number of new folks. In addition to all this, our visit to Vegas turned out quite successful for other reasons too - our "Identifying and Exploiting Windows Kernel Race Conditions via Memory Access Patterns" work was nominated and eventually awarded a Pwnie (in fact, two mascots) in the "Most Innovative Research" category. Woot!
While the subject of memory access pattern analysis or the more general kernel instrumentation was only mentioned briefly when we originally released the first slide deck and whitepaper, as we mostly focused on the exploitation of constrained local kernel race conditions back then, our most recent Black Hat "Bochspwn: Identifying 0-days via System-Wide Memory Access Pattern Analysis" talk discussed the specifics of how the control flow of different operating systems' kernels can be logged, examined or changed for the purpose of identifying various types of local vulnerabilities. Demos were presented live and are not available publicly (especially considering that one of them was a 0-day).
Slides: “Bochspwn: Identifying 0-days via System-Wide Memory Access Pattern Analysis” (5.26MB, PDF)
During the conference, we also open-sourced our Bochs kernel instrumentation toolkit (including both the CPU instrumentation modules and post-processing tools) under the new name of "kfetch-toolkit". The project is hosted on github (see https://github.com/j00ru/kfetch-toolkit) and is available for everyone to hack on under the Apache v2 license. Should you have any interesting results, concerns, questions or proposed patches, definitely drop us a line. We are looking forward to some meaningful, external contributions to the project. :-)
Last but not least, we have also explicitly mentioned that we would release all Bochspwn-generated logs that we had looked through and reported to corresponding vendors or deemed to be non-issues or non-exploitable. Below follows a moderately well explained list of reports, including information such as the original Bochspwn output, list of affected functions, our comments based on a (usually brief) investigation and the guest operating system and iteration number. Please note that a large number of Windows reports were assessed to be of a "NO FIX" class by Microsoft, and it might make sense to take another look at these and find out if the vendor didn't miss any obviously exploitable problems (unfortunately, we haven't had the time to perform a thorough analysis of each of the reports). Although a majority of the bugs were found in Windows, the Linux and BSD reports can certainly provide you with some interesting (yet not security-relevant) behavior and a fair dose of amusement. We hope you enjoy looking through the docs. Without further ado, here are the reports:
- Bulletin issues (fixed by Microsoft until 08/13/2013).
- Non-issues, non-exploitable problems and Denial of Service vulnerabilities.
- Unfixed issues as of 08/13/2013, "0-days".
All comments are more then welcome. Take care!