From: Eugen Leitl (eugen@leitl.org)
Date: Thu Jun 27 2002 - 02:29:21 MDT
On Thu, 27 Jun 2002, Alex Ramonsky wrote:
> Theory:
>
> Computers, such as we know them, cannot generate true random numbers.
This is out of date. You might examine the code for /dev/random.
Some chipsets incorporate high-bandwidth sources of entropy. It is only a
matter of time until such machinery migrates into CPUs. Its chiefest use
is defeating known state attack by mixing in true entropy into
pseudorandomness.
> Computer "random numbers" are in fact pseudo-random. A random number
Question is, can you tell the difference? The world might be completely
deterministic at the bottom. All entropy you observe could be pseudo at
heart.
> "seed" (a hidden variable) cycles through N different values, for some
> large value N. Even "cryptographically strong" random numbers follow the
> same system, the only difference being that N is hideously large.
So hideously large that you'll never observe a cycle.
> A random number sequence is obtained by permuting the seed, and
> returning f(seed) for some function f whose range is (in general)
> smaller than N.
>
> With sufficient analysis, it is possible to examine a sequence of
> pseudo-random numbers and determine the pattern - to predict the next
> one. For instance, the C function rand() could be completely predicted
> if you studied a sequence of 2^32 such numbers.
The C function rand() SHA1 is not. 2^160 is a little teensy bit larger
than 2^32.
I can readily give you an family of (untested) scalable hashes with
arbitrary length. The reason we don't have them is that we currently don't
need them.
> _true_ random numbers have no such pattern. They are truly random. They
> are completely unpredictable, no matter how many of them we study. Or so
> we believe.
I challenge you to tell the difference of SHA1 fed into intself or output
of /dev/random fed through SHA1. After you've done it with 2^160 we'll
repeat this with 2^16000.
> Of course, should it turn out that the perceived randomness of, for
> example, sub-atomic decay, were in fact pseudo-random, then it would
> follow logically that we _must_ be in a computer simulation. Or so it
> seems to me.
No. It follow logically that the universe process is deterministic. The
meta basis of the process is completely opaque if you're a part of it.
If the process is under control of sentients, they would need to
deliberately reveal themselves, since you can't detect them. There would
be no way to verify their claims.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:15:02 MST