Hacker News new | past | comments | ask | show | jobs | submit login

You can get around this. A biased HRNG is still an RNG, and there are other sources of entropy on a system anyways, so a decent OS can make exploiting a bias in an RNG infeasible.



Do the RNGs for fixed hardware devices like hardware security modules typically mix entropy from several sources?


I assume that HSM as discrete hardware devices certainly use their hardware RNG only as one input to some /dev/random like CSPRNG. On the other hand I would also assume that single-chip "HSMs" (smartcards...) do not although I vaguely remember that for TPM (which is hardware-wise a smartcard with weird host interface) the RNG output is somehow dependent on state of attestation registers.


Sure, many HSMs/smartcards/tokens lack sophisticated RNGs. And NIST's SP800-90 has proven weak on this.


I'm probably biased because I spend good part of quarter by designing reasonably secure CSPRNG for smartcard chip without hardware RNG (and ended up exploiting essentially any cross-clockdomain communication as entropy source) and thus I assume that typical smardcard vendors don't care about that (too much work) while HSM vendors simply leverage infrastructure of whatever (RT)OS they use and probably harden that somewhat.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: