News for hackers. Stuff that matters.
Re:Surreptitiously Tampering with Computer Chips by tptacek on Monday September 16, 2013 @05:34PM (#6396056)
This serves as your daily reminder that modern CSPRNG designs don't rely on a single entropy source, but rather mix many sources with hashes or block ciphers into entropy pools.

It would indeed have been bad if Linux (insanely) had converted to relying entirely on RDRAND as a replacement for the devrandom kernel CSPRNG. But no OS I know has done anything like that.

Even at 32 bits of entropy (which is an extremely unlikely conspiracy theory, as Schneier says), mixing the output of RDRAND into the rest of the entropy available to a kernel would still be a marginal improvement. For instance, it would mitigate some concerns about cold-start entropy for solid-state machines without real entropy sources.

The fact that OS CSPRNGs were unlikely ever to have trusted RDRAND directly (even if wasn't backdoored, what if someone screwed up? That too is somewhat unlikely, given that Paul Kocher's Cryptography Research reviewed the Intel HWRNG. But still...), and the fact that it's such a high-profile place to put a hardware backdoor, leads me to conclude that this is an unlikely vector for hardware backdoors.

There are no comments on this comment yet.