Re: [tlug] /dev/random is truly random?

["emiddleton@??" <emiddleton@??>]

Stephen J. Turnbull wrote:
>  > Stephen J. Turnbull wrote:
>
>  > > "twisted GFSR generator" (no, I don't know what that is
>
> emiddleton@?? writes:
>
>  > http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/ARTICLES/ttgfsr7.pdf
>  > http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/ARTICLES/tgfsr3.pdf
>
> Hey, I can google.  I even know what a feedback shift register is.
> What I can't do is explain why a "twisted generalization" is useful in
> crypto applications in terms that are intelligible, let alone
> convincing, to TLUG (or even to myself[1]).  If you can, please do.
>   

0000000000000000000000000000000000000000000000 could be produced by a
truly random source but you don't want your monti-carlo tests to use that.

One of the theoretical requirements of a random number generator is that
it produce k-distributed[1] sequences.  From the link.

"a sequence is 1-distributed if every number it generates occurs equally
often; 2-distributed if every _pair_ of numbers occurs equally often;"

Look at the book you referenced for a more detailed coverage.  The
twisted GFSR generator produces a sequence that is more k-distributed
and thus more useful.

Edward

1. http://www.bkgm.com/rgb/rgb.cgi?view+591