Platform/GFX/WebGL-Security-2011-10-28
From MozillaWiki
- Patterns of WebGL exploits so far (Benoit)
- More details have been sent to security-group on June 29 and October 28.
- Pattern A (Relying on a particular GL state, forgetting that scripts can change it)
- June 29 review: bug 659349
- October 28 review: no new bugs in this category
- Pattern B (Mistake when keeping track of GL state)
- June 29 review: bug 656752, bug 665070
- October 28 review: bug 665070
- Pattern C (Timing attacks)
- June 29 review: bug 656277
- October 28 review: no new bugs in this category
- Pattern D (Driver bugs)
- Type 1: Can be worked around.
- June 29 review: bug 631420, bug 657201
- October 28 review: bug 665578, bug 658826, bug 684882, Template:Bug 675625, bug 674042
- Type 2: Can be blacklisted. Tons of examples.
- Type 3: Ones we ignore for now because extensive DOS mitigations not available
- Type 1: Can be worked around.
- Pattern E (Implementation Bugs that are not at all GL-specific)
- June 29 review: bug 648705 in our WebGL implementation; bug 665934 in ANGLE.
- October 28 review: bug 686398, bug 685793, bug 682335 in our WebGL implementation; bug 680840, bug 665936 in ANGLE.
Securtiy Discussion
- go over action items from previous meeting and evaluate progress
- conformance tests should be in a very good shape now: https://etherpad.mozilla.org/WebGL-Conf
- ARB_robustness support finally landing (GLX already landed, WGL has patch)
- go over security bugs we've had since previous meeting and see how they fit in the patterns we discussed.
- regarding the DOS issue, discuss whether the pace of progress around ARB_robustness is enough or we want to do more.
- things seem to be progressing positively at this point
- discuss state of fuzzing
- still need to make a push to integrate fuzzcases as tests.
- If conformance tests revealed some bugs, does it make sense to mutate conformance tests?
- Does it make sense to do this without a real driver (software rendering)?
- Recent versions of MESA (7.11) with the llvmpipe driver (or softpipe) but avoid the old 'swrast' driver
- performance tests