Applied Cryptography Questions

A couple questions about Applied Cryptography (Second Edition)

1. What on earth is a Neural Net?

2. Consider the Chinese Lottery attack vs a processor bank.

   ADVANTAGES
     Distributed computing, no easy bombing target.

   DISADVANTAGES
     39% inefficient
     Need specialized hardware [for speed] or transmission of alogrithms
       for any specific alogrithm
     Not on all the time
     If value(crack) * cracks/chip > cost of chip, then why not have
       the gov buy the chips? Its probably cheaper that way.

3. How does one cryptoanalyize a Feistel Network? I've read a bit
   on this and it doesn't seem that the papers are consistant,
   for example, one claims that 2^32 chosen plaintext is unreasonable
   for a 64-bit block cipher whereas DES's cryptanalysis requires
   something like 2^42.

4. Does there exist an n such that a keyspace of 2^n is trivial to crack
   [ie a matter of miniutes] on a PC but is difficult to crack for a 
   big commercial company or even a major government? [this question
   relates to Merkle's puzzles, when a 40-bit key seems a bit weak.]

5. Is there an errata yet?

> 1. What on earth is a Neural Net?

This sounds like a troll, since we recently had a thread about the possible
use of neural nets in cryptographic applications, and the precise text used
in AC2 to discuss this matter. Very roughly, a neural net is an abstract 
model of a network of neurons in a brain, used in artificial intelligence as
a model of computation. I imagine you could easily have found more complete
information by doing some trivial net search for "neural net".

> 2. Consider the Chinese Lottery attack vs a processor bank.
> 
>    ADVANTAGES
>      Distributed computing, no easy bombing target.
> 
>    DISADVANTAGES
>      39% inefficient
>      Need specialized hardware [for speed] or transmission of alogrithms
>        for any specific alogrithm
>      Not on all the time
>      If value(crack) * cracks/chip > cost of chip, then why not have
>        the gov buy the chips? Its probably cheaper that way.

Well, is it in fact "probably cheaper" ?  If value(crack) > cost(chip) >
cost(attack) then the attack makes more sense....

> 4. Does there exist an n such that a keyspace of 2^n is trivial to crack
>    [ie a matter of miniutes] on a PC but is difficult to crack for a 
>    big commercial company or even a major government? [this question
>    relates to Merkle's puzzles, when a 40-bit key seems a bit weak.]
I think you need to explain more for this question to make sense. Obviously
companies (even big ones) and governments (even major ones) could buy and use
PCs themselves, in the event that conferred some cryptanalytic advantage.
ObLycos: Hey, better still, let's put the archives in an export-controlled
directory. We don't want too many outsiders to hear about the privacy
advantages of strong cryptography, now do we ?
ObAlices: <yawn>
-Futplex <futplex@pseudonym.com>
"I look pretty young but I'm just backdated" -The Who