Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site wdl1.UUCP
Path: utzoo!linus!philabs!cmcl2!seismo!hao!hplabs!hpda!fortune!wdl1!jbn
From: jbn@wdl1.UUCP
Newsgroups: net.crypt
Subject: Re: Why no hardware random numbers?
Message-ID: <331@wdl1.UUCP>
Date: Wed, 13-Mar-85 20:13:39 EST
Article-I.D.: wdl1.331
Posted: Wed Mar 13 20:13:39 1985
Date-Received: Sun, 17-Mar-85 20:57:10 EST
Sender: notes@wdl1.UUCP
Organization: Ford Aerospace, Western Development Laboratories
Lines: 32
Nf-ID: #R:utcsri:-86800:wdl1:28900005:000:1299
Nf-From: wdl1!jbn    Mar 13 15:43:00 1985


     One way to build a hardware random number generator is to
get a small radiation source and detector and use the pulses to drive a
counter.  Then use the following algorithm to generate one random bit.

	1.  Count pulses for a time T.  Call this count A.
	2.  Count pulses for a time T.  Call this count B.
	3.  If A=B, start over.
	4.  If A>B, emit a 1, else emit a 0.

     If T is too small, no bits will emerge, but the randomness of the
generator is not affected by the choice of T.  

     A friend of mine who was developing commodity speculation software
once reached the point of designing an S-100 bus random number generator
using the above technique, but he got out of the S-100 card business before
building it.  His problem was that he tested his commodity speculation programs
against a random walk; the ideas is that if they practice good money management
they should not lose badly against a random walk.  They were winning.  This
is impossible with a true random walk; however, his analysis programs were
picking up patterns in the random number generator and were able to extract
some pattern from the noise.

     True random number generators are available commercially for key
generation, from

	HAGELIN CRYPTO AG
	P.O. Box A-163
	CH-6301 Zug/Switzerland

						John Nagle