[boost] Re: Proposal: random number library

From: Reid Sweatman (borderland_at_[hidden])
Date: 2000-02-23 14:57:56

I'm familiar with the Yarrow paper.  Granted that it suggests restricted
algorithms for hashing; I was under the impression, though, that the
proposed Boost library was not intended for cryptographic purposes (correct
me if I'm wrong).  In such a case, it would be a simple matter to substitute
something less hobbled by silly laws for the hash.  In addition, the Boost
library could be used as a high-end shell, with the low-level algorithm of
choice inserted by the user, where something crypto-ish was required.

As for games, you might be surprised at what is commonly used.  The last
commercial game I worked on used R250.  The Mersenne Twister is the darling
of the gaming avant garde right now.  One curious point about game PRNG's is
that you generally *want* them to be exactly repeatable, for implementing
things like simulation recorders, and so on.  That means that Yarrow, or
anything else based on an entropy pool, or direct sampling of a
supposedly-random source, such as semiconductor junction noise, isn't of
much use in games.

> -----Original Message-----
> From: jmaurer_at_[hidden] [mailto:jmaurer_at_[hidden]]On
> Behalf Of Jens Maurer
> Sent: Tuesday, February 22, 2000 11:29 AM
> To: boost_at_[hidden]
> Subject: [boost] Re: Proposal: random number library
>
> There is a difference in quality between a PRNG for simulations and one
> for cryptographic applications.  For simulations, you ask questions such
> as "does this PRNG have good statistical properties, i.e. does it pass
> the D.E. Knuth tests?" and "is it fast (enough)?".
> However, for cryptographic applications, you must make sure that nobody
> can guess your random numbers even if he knows some parts of the sequence
> of your PRNG.  Execution speed is secondary.  Read the paper about
> Yarrow-160 on www.counterpane.com for a cryptographer's approach to PRNGs.
> In this paper, the authors suggest using SHA-1 and Triple-DES for
> hashing.  Triple-DES certainly qualifies as an encryption algorithm.
> Besides, it's not only the US ITAR regulations. In France, for example,
> use of cryptography requires an individual license from the government.
> Look at your favourite Netscape mirror for the special crypto-disabled
> "France" version.