[Cryptography] The ultimate random source
Joseph Ashwood
ashwood at msn.com
Fri Feb 21 07:52:13 EST 2014
-----Original Message-----
From: Bill Frantz
Subject: Re: [Cryptography] The ultimate random source
> I feel better with a real image that can test that the camera is still
> working.
Just make sure there is absolutely 0 post-processing that destroys entropy,
this is a lot harder than it sounds. This is exactly why I spent so much of
my initial email about the debayering/demosaicing process, the popular
processes destroy a significant amount of the entropy. I know it may seem
irrelevant at first, but any system that starts by throwing away 2/3 of the
data in a deterministic process, and then proceeds to process the remaining
data to smooth out any strange bumps throws away most of the entropy.
> We get 6 ** 600 possibilities and we haven't even considered small changes
> in position of each M&M. 6**600 overflows my HP15 calculator and makes me
> feel comfortable seeding a PRNG.
Certainly the 6^600 starting states sounds like a lot, but even assuming
that it is somehow perfect, that is not much more than 1200 bits of entropy.
Harvesting this 1200 bits of entropy 512 bits at a time is only 2 full
samples. This is why the continued introduction of entropy matters, and why
the general lack of entropy introduction matters even more.
Joe
More information about the cryptography
mailing list