Re: Verification of Random Number Generators

[droelke@rdxsunhost.aud.alcatel.com (Daniel R. Oelke)]

> 
> Just an idle thought: it might be possible to do a probabalistic
> verification of a RNG by sampling it over some number of samples, and
> statistically analyzing the sample space. This would be analysis under the
> model of "RNG as black box" as opposed to (or rather, if you're smart, in
> addition to) code inspection & review. Any statisticians among us?
> 

In a word - no.

In a longer description - you can black box test the *pseudo-RNG in
one of two ways.  

One, input a key, and then keep asking for numbers.  The numbers 
should be statistically distributed, and should have a very long 
repeat cycle.  (i.e. basic "good" pseudo-random number generator
theory)

Two, input different keys, and then ask for a number.  Again,
the numbers should be statistically distributed, even for small
changes in the input numbers.  

What *both* of those methods don't test, is the key(seed) generation,
which is what was attacked on Netscape.  The seed number
must be truely random, and unknowable to the attacker, otherwise
they can duplicate the random number generator.  Remember that
all random number generators in software are really pseudo-random
number generators, and rely on having seed value.

BTW: In some ways, I feel sorry for Netscape being nailed by everyone
including the press like this.  BUT - they must also accept that
with being popular, that they will be subject to intense 
scrutiny.  Besides - if you are going to attack a cryptosystem, why
not get your 15 minutes of fame out of deal while you are at it ;-)

Dan
------------------------------------------------------------------
Dan Oelke                                  Alcatel Network Systems
droelke@aud.alcatel.com                             Richardson, TX