Atmospheric noise & fair coin flipping

Optimizzin Al-gorithym al at qaeda.org
Sun Jul 14 09:17:34 PDT 2002


At 05:45 AM 7/14/02 -0700, gfgs pedo wrote:
>it is said that atmospheric noise is random but how
>can we say for sure.

Physics, chaos, the growth of initial uncertainty as systems evolve,
energy/time required to make measurements to arbitrary precision.


>what if the parameters giverning atmospheric noise
>vary frm time 2 time.

The rules of physics are those that don't change from time to time, or
place to place.
Certainly the e.g., wind speed does.

>so can we say atmospheric noise is random or a coin
>flipping is random-only because it passes die hard
>test or other randomness tests-which is an indicator
>of randomness with the current defenition of
>parameters in determing randomness?

No, since 'anything through a whitener passes' these tests.
The integers (0, 1, 2..) fed into DES will pass.
(Equivalently) A low-entropy source fed into a hash will pass.

[Historical note: this is why Intel should make its raw RNG
data available in chips with whitened-output RNG functions]

To have a true RNG, You *must* have a physical understanding of the
source
of entropy whence you distill the pure bits (whether or not
you feed it into a whitener after distillation).  Precisely
because a 'black box' may be a deterministic (if you know
the secret) PRNG.  By 'distill' I mean reduce N bits to M,
N > M, in such a way as to increase the entropy of the
resultant M bits.


>is there truly random or that we can say with certain
>degre of confidence that they are nearly random as all
>current evidence poits so.

'Random' should be taken to mean 'ignorant of'.  It suffices
that we (and our adversary) are ignorant of the detailed conditions
inside a noise diode, unstable atomic nucleus, atmospheric
(or FM radio) noise receiver, etc.  Philosophical discussions about
'true
randomness' ("Is there a deeper/smaller level of description in
which apparently-random events are based or emerge from?")
are beyond the scope of this rant.





More information about the cypherpunks-legacy mailing list