[117590] in Cypherpunks
re: otp
daemon@ATHENA.MIT.EDU (Anonymous)
Mon Sep 6 19:13:29 1999
Date: Tue, 7 Sep 1999 00:49:20 +0200 (CEST)
Message-Id: <199909062249.AAA12963@mail.replay.com>
From: Anonymous <nobody@replay.com>
To: cypherpunks@cyberpass.net
Reply-To: Anonymous <nobody@replay.com>
> You need audio volume: people
> who record ambient noise, stereo-differences, etc.,
> aren't getting the entropy-per-second they could by
> e.g., placing the microphone near a spray nozzle on 'mist',
> or (if you live in a dry climate) by using the
> noise in a detuned radio, played loud.
No, that's stupid! Just use the basic thermal noise available from an
appropriately designed circuit. There is no reason to build an audio
input and then hook it to a sprinkler, for crying out loud! If you want
that kind of Rube Goldberg appartus then at least have the decency to
include a snoring man in the system. But really, there is plenty of
randomness in the basic electronics. This is what the Intel chip uses.
Specifically, they have a high frequency oscillator being sampled by a
low frequency oscillator. The period of the sampling clock is modulated
by the differential thermal noise in a pair of resistors. This makes the
phase at which the HF clock is measured become almost completely random,
to many significant digits of precision.
> BTW, I don't think the P-III lets you look at the
> raw bits out of the RNG, does it? Some people like
> to brew their own... and monitor the brewing process.
There are several ways of looking at this question.
First, what is meant by "raw"? The RNG internally has two stages. The
first is described above, a high-frequency clock which is sampled at
random intervals. This is then passed through a second stage which
removes any bias (say, caused by the fact that the duty cycle of the
HF clock is not perfectly uniform). This second stage uses a variant
on the widely-known technique of looking for 1-0 or 0-1 transitions and
only outputting when those are seen.
The output of the chip is the result of this second stage. So, the answer
to your question is no, if by "raw bits" you meant that you wanted to see
them before they went through the bias removing circuit.
Intel has not published the spec for accessing the bits from the chip.
They sell a library which implements a standard API for reading the bits.
The library further whitens them by passing them through a SHA-1 hash
function.
At present, if by "raw bits" you mean that you just want the bits from
the chip without going through Intel's library, the answer is no, the
specs have not been published to permit this. However it is likely
that before long the data will be made public, by reverse engineering
the driver if nothing else.
So within a few months the answer to your question should be yes, it will
be possible to read the raw bits from the RNG.
