1993-05-09 - RNG from zener is easy.

Header Data

From: William Oldacre <76114.2307@CompuServe.COM>
To: <cypherpunks@toad.com>
Message Hash: e757ff7355b8317e76725ec9bd42246d868faf950c4b50c07dad0908eee9d9cb
Message ID: <93050917021576114.2307_BHA35-1@CompuServe.COM>
Reply To: _N/A

UTC Datetime: 1993-05-09 17:07:36 UTC
Raw Date: Sun, 9 May 93 10:07:36 PDT

Raw message

From: William Oldacre <76114.2307@CompuServe.COM>
Date: Sun, 9 May 93 10:07:36 PDT
To: <cypherpunks@toad.com>
Subject: RNG from zener is easy.
Message-ID: <930509170215_76114.2307_BHA35-1@CompuServe.COM>
MIME-Version: 1.0
Content-Type: text/plain

To: >Internet:cypherpunks@toad.com

This posting is in response to Liam's earlier message on creating a 
random number generator using a zener diode.  I breadboarded such a 
circuit over year ago and it worked fine.  Zener diodes, operated in 
their zener region with a reverse voltage generate electrical noise 
much like a "noise diode".  This can be amplified and shaped using a 
single quad op amp package to create what appears to be a very random 
data stream.  Best way to use this would be to shift it into an eight 
bit latch before reading the parallel port.  Depending on design, 
it's operation is much (MUCH) faster than a pseudo random number 
generator written in software.  It can fill a floppy disk so fast 
that the main limitation  is the writing speed of the disk drive.

A simple XOR with the message text is enough to create an unbreakable 
cyphertext.  Problem is, that both parties have to have exact copies 
of the random numbers since they cannot be regenerated as with pseudo 
random numbers using a key.  What is needed is a way to encrypt the 
random numbers so that they can be sent to someone else to use.  This 
would be hard to decode, because it would be difficult to know when 
success had been achieved.  If it is being tested against a message 
text, the loss of one byte shifts everything to that the text becomes 
gibberish again.

I'm glad Liam brought this up.  It is a good way to produce an 
absolutely unbreakable code.  This, providing the design is good and 
no patterns are allowed to appear (introduced by amplifier 
oscillation or nearby noise sources) in the generated numbers.  If 
there is any interest out there perhaps we (Liam or I) could produce 
a few for testing.