bram (bram@gawth.com)
Tue, 14 Jul 1998 21:38:50 -0700 (PDT)
On Tue, 14 Jul 1998, David Wagner wrote:
> it would be better to prevent iterative-guessing attacks. The
> typical countermeasure is to save up a number of entropy samples,
> and only mix them into the pool once their combined entropy exceeds
> some threshold (perhaps 64--160 bits, according to taste).
Fortunately, the way most hashes work it's possible to compute the hash of
the concatenation of a whole bunch of separate bitstrings without having
to keep them all in memory till the end.
I earlier suggested using a separate PRNG to pool data, which given the
above insight is completely unnecessary.
-Bram
(who never rewrites posts three times because the first couple things he
wrote were either wrong or embarassingly stupid.)
The following archive was created by hippie-mail 7.98617-22 on Fri Aug 21 1998 - 17:20:24 ADT