building a true RNG

[Sampo Syreeni]

On 2002-07-28, Sampo Syreeni uttered to David Wagner:

[Answering to my own mail. Sorry.]

>and discard every 1/(p(x)-1/256)'th sample with value x.

Actually the pedantic solution would be to put an arithmetic
compressor/coder between the input and output, using the best model we've
got. That still leaves model adaptation to be dealt with, but if we
discard a sufficient number of output bits at start (estimable from the
model), we *will* end up with (very nearly) flat statistics on the output.
Asymptotic optimality and all that... (The qualification comes from
limited precision arithmetic.)
-- 
Sampo Syreeni, aka decoy - mailto:decoy at iki.fi, tel:+358-50-5756111
student/math+cs/helsinki university, http://www.iki.fi/~decoy/front
openpgp: 050985C2/025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2


---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at wasabisystems.com