3 ms·
It's a little confused. Stagefright is just a library containing the exploit. The component being (presumably) exploited is the mediaserver process. This is
by ajross 11y ago
It's a little confused. Stagefright is just a library containing the exploit. The component being (presumably) exploited is the mediaserver process. This is where almost all video decode/encode handling happens in Android (and other related stuff, like audio and camera).
And yes, the mediaserver has internet access so that it can fetch and decode streams on its own without forcing all I/O to go through a user process. Basically it's a performance optimization, though indeed probably one necessitated by having to wall off all that DRM handling in the first place.
That said, mediaserver, while it has lots of access to things that Android apps don't (e.g. open file descriptors to kernel vendor-specific drivers with lets-just-say-questionable security practices), is not a root process. And it's reasonably sandboxed in recent Android versions using selinux. An exploit into it may not be quite as device-cracking as is being claimed.
Still, not a good thing at all.
- Florin_Andrei 11y ago> And it's reasonably sandboxed in recent Android versions using selinux. Is it safe to assume that this attack vector is not a concern with Android 5.1? (or later)
- ajross 11y agoCertainly not, because the plausible kernel bugs in those open driver descriptors (I'm serious: vendor media drivers are not pretty things) would obviously not be subject to selinux controls. But at least there is some sandboxing in place; the security architecture was designed to admit the possibility of a breach of the mediaserver.
- fluidcruft 11y agoAccording to the reddit discussion, 5.1.1 is demonstrated to be vulnerable (so it affects everything 2.2 and later).
- ajross 11y agoLink to that reddit thread? AFAICT, all current claims are for a crash exploit into libstagefright, not a breakout of the mediaserver to (for example) persistently write a payload/rootkit.