Re: R: True random numbers wanted

From: Charles Hixson (charleshixsn@earthlink.net)
Date: Sun Sep 01 2002 - 21:48:43 MDT


On Sunday 01 September 2002 11:35, Eugen Leitl wrote:
> On Sun, 1 Sep 2002, Charles Hixson wrote:
> > Well, you could take a feed from a geiger counter. That's pretty
>
> I don't think it's any more random than noise from a turbulent airstream.
> I think the larger the region the noise is coming from, the better.
>...
> I don't want a distribution of values, just entropy (which I might or
> might not destill). All the rest is cryptohash's work.
>
In that case, write a function that accumulates based on a function of how
long since you last struck a key. Let it run in the background and flip bits
in an unpredictable way as you use your computer (perhaps: on key down, take
the number of ms since the computer was last turned on, mod 32 [or 64], and
flip that bit.)

I'm not quite sure what standard you are using for random, but that's a way
that works pretty well if all you need is unpredictable values at a low rate.
You could also save values from this after every, say 128 keystrokes, if you
needed a pad of such numbers. (I believe that Linux, and perhaps other
*nixes, have a device that already does this kind of thing, called
/dev/random, but I've never needed to use it so I haven't investigated. This
is considered pseudo-random rather than actually random, but it's random in
the sense of unpredictable.)



This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:16:38 MST