Re: [Demexp-dev] VIA x86-like processors with cryptographic facilities

[David MENTRE]

Hello Brian,

Brian Hurt <address@hidden> writes:

> I wouldn't trust the random number generator.  As a source of randomness, 
> yes- but I'd wash them through a cryptographic hash to wash out any 
> biases.  It's really easy for biases to creep into hardware RNGs.

Yes, you have raised a good point. Anyway, before using such generator,
I would use tests on generator to evaluate its randomness.

And BTW, it is possible that the random generator is not a proper
generator based on a physical device like thermal noise. But, in that
case, cryptographically hashing them wouldn't help: if the source of the
cryptographic hash is predictable, then the hashed value is
predictable. That's why having a real source of randomness is important.

I said it was a solution to consider, I didn't said it was THE
solution. :) And we are far from being at that step yet. :(

Yours,
d. -- probably not qualified to speak of cryptography
-- 
 David MENTRÉ <address@hidden>