From: Simon Spero <ses@tipper.oit.unc.edu>
Date: Thu, 5 Oct 95 21:18:29 PDT
To: Norman Hardy <norm@netcom.com>
Subject: Re: Simple Hardware RNG Idea
In-Reply-To: <ac9a42dc01021004b46d@DialupEudora>
Message-ID: <Pine.SOL.3.91.951005210923.11085D-100000@chivalry>
MIME-Version: 1.0
Content-Type: text/plain


On Thu, 5 Oct 1995, Norman Hardy wrote:

> Simon Spero writes:
> ....
> > What about a beam of high intensity ionising radiation aimed at the
> > detector?
> ....
> You presumably use the oddness of the count for your random bit in some
> predetermined time interval. External radiation can change, but not bias
> the parity. If the counter saturates, the counter may be biased towards one

Hmmm. But isn't this method slightly biased? If the probability of  N 
events < the probability of N+1 events, wouldn't you need a large number 
of events per bit to make the bias insignificant? 

The measurement I was thinking of (which would have been susceptible to 
the external attack) was to measure the interval between events, and 
convert that to a uniform distribution. That's probably trying to get too 
many random bits per event, but does let you use much lower level sources.

BTW, I was just having dinner with a bunch of guys from HP labs in 
Bristol, UK, and the subject of hardware RNGs came up. The idea of 
Strontium-90  as the next computer consumable has a certain appeal- the 
designating brandname is "Omnisource".

 "Are you having trouble scintillating? With Omnisource, you can scin till
  dawn!" 

Simon