Re: coding and nnet's

[Bill Stewart <stewarts@ix.netcom.com>]

At 02:17 PM 11/10/95 -0500, Atp3000@aol.com wrote:
>I just resently got interested in coding and cryptography, and I was
>wondering if you could suggest 3 or 4 papers that adresses the aplication of
>neural networks in cryptography.

Schneier's 2nd edition says "Neural nets aren't terribly useful for
cryptography,
primarily because of the shape of the solution space.  Neural nets work best for
problems that have a continuity of solutions, some better than others.  
This allows a neural net to learn, proposing better and better solutions as
it does.
Breaking an algorithm provides for very little in the way of learning
opportunities:
You either recover the key or you don't. (At least this is true if the
algorithm is
any good.)  Neural nets work well in structured environments when there is
something 
to learn, but not in the high-entropy, seemingly random world of cryptography."
And he doesn't give any references.

That's been my opinion of the issue as well; I looked into it a bit when I was
doing a project with the neural net folks back at Bell Labs, partly because
neural net chips typically have lots and lots of parallel bit-sized horsepower.
Unfortunately, the horsepower isn't arranged in ways that are very useful for
crypto;  adding together a large bunch of short chunks of data (maybe using
floating
point addition) and thresholding the sum isn't the right thing to do with
highly discontinuous functions.  You could take a similar chip design and
connect
the pieces together differently to make a brute-force searcher, i.e. take a gate
array and wire it to do crypto-like calculations, but the neural net stuff
doesn't do that very efficiently.
#--
#				Thanks;  Bill
# Bill Stewart, Freelance Information Architect, stewarts@ix.netcom.com
# Phone +1-510-247-0663 Pager/Voicemail 1-408-787-1281