[Cryptography] Dieharder & symmetric cryptosystems

Michel Arboi michel.arboi at gmail.com
Wed Mar 23 10:20:22 EDT 2022


Le sam. 19 mars 2022 à 05:54, Richard T. Carback III <rick at carback.us> a écrit :
> I concur the issue is with the tests... all of the recent changes are net-positive from a cryptanalysis security perspective (unless, of course, there’s a bug).

When I encountered these strange results years ago, I excluded this hypothesis:
1. I go the same behaviour with "dieharder -g 501 -a" and "cat
/dev/urandom | dieharder -g200 -a"
There has been a bug in Dieharder on 64 bits systems which made the
500 & 501 (random & urandom) fail badly. It has been fixed years ago.
The 200 "generator" reads random bytes from stdin.
2. At that time, I did not have these WEAK results with 500 (random),
only with 501 (urandom). I did not have them with a bogus "SHA1 RNG"
either (hash a incrementing counter and use the result as "random
bits")
I never found any sane explanation at the time.

The recent changes align /dev/random results with /dev/urandom


Now, there is probably a problem with test 201 (RGB Generalized
Minimum Distance Test). I'm looking at the other tests that "often
fail"...


More information about the cryptography mailing list