Re: genetic algorithms for crypto analysis

It has been noted in this thread that a good crypto algorithm would
  require an attacker to locate a single spike in a problem space, rather
  than having to climb a hill (which is, of course, much easier).
  
  I recall reading (I think in Sci. Am.) that a theory under investigation
  now as to why nature has sexual reproduction as part of its repertoire
  is that this gives a solution-seeking population a better opportunity to
  located spikey solutions.
  
  From the point of view of genetic algorithms, sexual reproduction means
  that each offspring must be generated from two members of the existing
  population, each of which contributes half the information needed to
  generate the offspring.  In theory, this maintains a population that
  is spread over a wider terrain, and is thus more likely to find the
  spike.
  
  I don't know if such a strategy would help at all in crypto analysis,
  or whether any genetic algorithm programs currently in use employ this
  strategy.
  __
  |         (V)              |  "Tiger gotta hunt.  Bird gotta fly.
  |   (^    (`>              |   Man gotta sit and wonder why, why, why.
  |  ((\\__/ )               |   Tiger gotta sleep.  Bird gotta land.
  |  (\\<   )   der Nethahn  |   Man gotta tell himself he understand."
  |    \<  )                 |  
  |     ( /                  |                Kurt Vonnegut Jr.
  |      |                   |  
  |      ^                   |

HAHN@lds.loral.com:
> [makes excellent point that given sexual reproduction, evolution
>  does not need continuous search space]
>   I don't know if such a strategy would help at all in crypto analysis,
>   or whether any genetic algorithm programs currently in use employ this
>   strategy.

Sexual reproduction (aka string crossover) is the fundamental attribute 
of GAs that distinguish them from hill-climbing algorithms; it has been in 
all GAs from their invention.  One of original works on the subject is 
now out in reprint: John Holland's _Adaptation in Natural and Artificial 
Systems_, MIT Press.  

Crossover doesn't allow magic teleportation directly to the
needle in the search space haystack.  GA leaps over gaps where the 
"crossover Hamming distance" is small, but the space need not be continuous.
Cryptanalysis where one can gain clues, partial solutions, etc. and
compose these into better solutions, might be amenable to GA.
If you can say "solution A is better than solution B" with an 
algorithm, it's a good candidate for solving with GA or GP (genetic 
programming, which works on trees instead of strings).

Nick Szabo				szabo@netcom.com