Re: new /dev/random
From: Paul Rubin (//phr.cx_at_NOSPAM.invalid)
Date: 10 Oct 2004 13:12:55 -0700
email@example.com (S. Vinder) writes:
> I assumed D() was used as a mixing function, whose purpose was to
> conceal possible patterns and correlations in data gathered from
> sources like mouse movements.
Yes. The data from the mouse movements is random but comes from an
unknown and presumably non-uniform probability distribution. Could
there be an unlucky (i.e. malicious) choice of distribution that still
gives the required amount of entropy in the mixer input, but results
in predictable mixer output? We're trying to specify mixer
characteristics that stop that from happening.