No it bloody doesn't.

[ciphergoth]

Mathematicians who know fuck all about crypto are fond of saying that their latest discovery might have crypto applications.

http://news.bbc.co.uk/1/hi/sci/tech/2146295.stm

In this case, as usual, it doesn't.

Why is it crypto, of all fields, that attracts this idea that you don't have to know a damn thing about it to innovate in it? All fields get crackpots, but even crackpots have a vision that there are people employed to do some research in this field already, whereas there seem to be an endless supply of people who act as if they are the first to think really hard about encryption.

Update: Whoops, I spoke too soon. It turns out that Carl Pomerance among others is involved in this research, so I guess it is legit. I'm surprised.

Comments

[babysimon]

It's the distinction between random and pseudo-random. Unless you are very careful, any pseudorandom number generating function you design can be reversed easily to derive the seed/key from a sequence of pseudorandom numbers generated from it.

Of course, a truly random sequence would be useless for crypto as you could not reconstruct it at the other end. So you need a pseudorandom generator function which is also a one-way function. This is a hard problem.

Of course, IANAC.

[ciphergoth.livejournal.com]

This is basically correct. This "normality" property they've proven is not enough for a generator to be useful - it must be "computationally indistinguishable" from a true random number generator to someone who hasn't got the seed. We've got lots of generators that we believe are indistinguishable, most of which are *far* faster than fetching digits of Pi with the Bailey-Plouffe algorithm, and plenty that have much better theoretical properties than BP.

A quick email exchange with David Bailey has done nothing to allay my suspicions and everything to confirm them...