Re: Evolving algorithm for faster brute force key searches?

[Adam Shostack <adam@homeport.org>]

Timothy C. May wrote:

| With some weak ciphers, this might work. I think Schneier makes some
| comments about who's looked at this. But weak ciphers are not too
| interesting.

	At the most recent Crypto, someone mentioned that FEAL is
useful because just about any new attack you can think of works well
against it.  I think it was Susan Langford.

	Weak systems are thus useful for research and training
purposes.  I suspect Tim is on the money with a genetic algorithim
having a flat `fitness landscape,' but there may be something that a
human misses which an evolved algorithim finds.

	Also, it may be possible to evolve something against a
reduced round version of a cipher (using a training space that is not
flat) that will still work better than brute force against a full
system.  If you have cycles to spare, it might be an interesting
avenue of research.

Adam


-- 
"It is seldom that liberty of any kind is lost all at once."
					               -Hume