From: HO  JUNYA <hojunya@ecf.toronto.edu>
Date: Wed, 19 Jan 94 10:08:56 PST
To: cypherwonks@lists.eunet.fi
Subject: Questions about hardware RNG design and implementing RSA/DES/IDEA in hardware
Message-ID: <94Jan19.130755edt.4305@cannon.ecf.toronto.edu>
MIME-Version: 1.0
Content-Type: text/plain



(I apologize for the crossposting; I feel that this is relevant to
the newsgroups and mailing lists  that I've posted this to)

Hi, we're trying to design and build a hardware random number generator
for use in key generation for secret key cryptosystems (likely to
be at least a one time pad, perhaps DES or IDEA), as part of a
hardware design project.  We'll be putting the transmitter and
receiver on expansion cards for ibm-compatible pc's, and hopefully
be able to keep as much of the work as possible in hardware.

My question is, I am trying to come up with ways of building
this true RNG (random number generator) using a few simple components.
For my noise source, I will probably be using thermal noise in
electrical components.

I understand that noise has a Gaussian distribution  through
time (i assume time referring to the time interval between successive
signals of a certain power/amplitude?), and will be finding the
median of this distribution so as to sample the random binary bits
on either side of this median.  We will attempt to design something
which will automatically find the median in a given set of
conditions.  (I am told this is called anti-biasing?)

My questions are:

does anyone know of any actual implementations of this sort of RNG?  

can anyone suggest a good introductory reference to noise, random
behaviour or sampling from random sources?

what sort of hardware requirements do the DES, IDEA and RSA algorithms
have?  We will not be using IC's built for the DES or RSA encryption,
but will perhaps program a controller of some sort to do this.  I
am afraid that RSA (for the initial secret key exchange) will take
too long on an 8088 or lesser processor.  we're looking for bitrates
of about 1kbits/s from the RNG, since 75% of that will likely be
thrown away in the process of 'pairwise (something)'.  (if two 
consecutive bits are the same, throw them out, if they are different,
take the first one)


I would appreciate any comments, answers, concerns, &c

Junya
hojunya@ecf.toronto.edu