Security/Meetings/SecurityAssurance/2012-03-13
From MozillaWiki
< Security | Meetings | SecurityAssurance
- Time: (Weekly) Tuesday at 13:30 PM PDT / 16:30 PM EDT / 21:30 PM UTC.
- Place: Mozilla HQ, 3A-All Your Base (3rd Floor)
- Phone (US/Intl): 650 903 0800 x92 Conf: 95316#
- Phone (Toronto): 416 848 3114 x92 Conf: 95316#
- Phone (US): 800 707 2533 (pin 369) Conf: 95316#
Agenda
- Goals - 2 weeks left
- Q2 Goals
- Team
- Bugzilla Security Mail - https://wiki.mozilla.org/Security/Security_Bugs/EncryptedBugmail
- Pwn2Own Update
- Status updates (cutisk)
- Static Analysis
- https://security.etherpad.mozilla.org/Static-Analyzers (add stuff to it)
- Tag Merger update (curtisk)
- SecReview form update
Second half
- [decoder] ASan as a nightly and/or for hardened environments
- B2G permissions model (dchan, want to chat?)
Project Updates Below
B2G
- Discussions on the Permissions Model ongoing
- Developer Phone release 4/1/12 [1]
- Q1 Goals - complete review for developer phone
- Dependant on B2G hitting this milestone (currently 5/40 features are "done".)
- https://wiki.mozilla.org/B2G/Schedule_Roadmap
- Q2 Goals - product phone aiming for release by end of Q2 so:
- Complete all necessary security reviews
- Complete documentation of B2G security architecture (is this something we are responsible for? I guess not, but I think we want to be involved)
- Develop B2G fuzzing platform?
Automation Tools
- Debate on whether Marionette should land on Gecko between devs and code reviewers in bug 712643.
Fuzzing
- [gkw, decoder] m-c, IonMonkey fuzzing continuously underway
- decoder has a lot of asserts, gkw a lot of hard-to-reproduce-by-IM-devs GC bugs, Jesse a lot of general bugs
- [decoder] Continue mobile fuzzing goal from Q1
- In Q2: Deploy fuzzers compatible with ADBFuzz developed in Q1 (e.g. jsfunfuzz, domfuzz)
- Involves getting necessary hardware and process in place
- In Q2: Deploy fuzzers compatible with ADBFuzz developed in Q1 (e.g. jsfunfuzz, domfuzz)
- Move fuzzers to Releng hardware. jsfunfuzz harness migration underway.
- Non-mobile hardware already in-place.
- [gkw] Needed if we are to release funfuzz in April
- why?
- I wouldn't want to release old code from literally years ago only to overhaul them as they get released. Makes documentation easier (documenting the new code) as well.
- I'd like if releng committed to fixing https://bugzilla.mozilla.org/show_bug.cgi?id=628573 ;)
- They rank it P3.
- [gkw] Several Valgrind blockers for Mac OS X Snow Leopard & Lion fixed by sewardj
- [jesse] tweaking jsfunfuzz to find bugs like our pwn2own bug https://bugzilla.mozilla.org/show_bug.cgi?id=720079
- [gkw, decoder] m-c, IonMonkey fuzzing continuously underway
ASan
- http://blog.mozilla.com/decoder/2012/01/27/trying-new-code-analysis-techniques/
- http://blog.mozilla.com/decoder/2012/03/09/update-on-address-sanitizer/
- Q2 goal: get builds from RelEng
- Get regular testing of Firefox setup
- Integrate with our crash reporting so we can receive ASan crash reports once we decide to have more people testing ASan builds (e.g. a special nightly).
- This could be great for troubleshooting. "Hitting an intermittent crash? Try an ASan build and see if it becomes a reliable crash."
- ASan builds as a "hardened Firefox" for paranoid users?
- For what kinds of does ASan really prevent exploitation, as opposed to just requiring attackers to change offsets or do more allocations?
- Use-after-free as well as stack+heap buffer overruns should be covered
- How many people are willing to take a 3x slowdown for it?
- How would this compare to using debug builds with (at least some) assertions fatal?
- Maybe we should enable the nsTArray and nsPresArena and compartment assertions for these builds, but not all assertions.
- Target audience size not as big, asan builds likely Linux/Mac-only
- For what kinds of does ASan really prevent exploitation, as opposed to just requiring attackers to change offsets or do more allocations?