Path: cactus.org!milano!cs.utexas.edu!wupost!micro-heart-of-gold.mit.edu!uw-
+     beaver!ubc-cs!unixg.ubc.ca!unruh
From: unruh@physics.ubc.ca (William Unruh [Unruh])
Newsgroups: sci.crypt,comp.compression

Subject: Re: Source for _real_ random numbers
Message-ID: 
Date: 18 Mar 92 00:30:50 GMT
References: <1992Mar16.153445.5771@cs.dal.ca> <92Mar16.131831edt.203@neuron.
+           ai.toronto.edu>  <1992Mar17.185517.2014@athena.mit.edu>
Sender: news@unixg.ubc.ca (Usenet News Maintenance)
Distribution: usa
Organization: University of British Columbia, Vancouver, B.C., Canada
Lines: 18
Xref: cactus.org sci.crypt:5588 comp.compression:2553
Nntp-Posting-Host: physics.ubc.ca

jfc@athena.mit.edu (John F Carr) writes:


>Someone who used to work here used a disk drive to generate random numbers.
>He measured the time required to read a disk block and used the variation in
>that time (after much preprocessing) as a random number source.  A large
>part of the variation in disk rotation speed is caused by air turbulence, so
>there is some physical as well as observational evidence that these numbers
>are random.

 In fact I think one has to be very careful with using physically
 generated random numbers. Although the underlying process might well be
 random, one is introducing all kinds of measuring instruments in
 between you and the underlying process. Those instruments could well
 introduce biases and correlations. One approach might be to use the
 physically generated random numbers as seeds to a deterministic
 "randomiser" (eg encrypt it with DES say) to get rid of the low order
 biases or correllations of the physical generator.