From: Phil Karn <karn@qualcomm.com>
Date: Thu, 23 Jun 94 15:15:39 PDT
To: ravage@bga.com
Subject: Re: your mail
In-Reply-To: <199406221812.NAA21717@zoom.bga.com>
Message-ID: <199406232210.PAA16536@servo.qualcomm.com>
MIME-Version: 1.0
Content-Type: text/plain


>This is a bad idea, the computer it self will generate clocking noise which
>will appear in the noise and destroy the randomness. The standard, and even
>most high-end, recievers don't have the shielding to prevent this sort of
>intrussion. Heck, that digital clock on your desk (and possibly your wrist if
>close enough) will cause problems as well. 

In my ham experience, almost all of the computer noise you might pick
up in such a receiver comes through the antenna. If you disconnect the
antenna, what remains is almost entirely thermal noise from the
receiver front end. And even if there were some computer noise mixed
in, MD5 hashing the output to concentrate the entropy should satisfy
even the most paranoid.

It's advisable to replace the antenna with a dummy load (matched
resistive terminator) to ensure that the front end remains stable; if
the receiver is sensitive enough you also pick up the thermal noise
generated by the terminator itself. (BTW, you can easily demonstrate
thermal noise with a *good* low-noise preamp and a linear, e.g., SSB
or AM -- not FM, receiver this way.  Turn off the receiver AGC and
dunk the terminator in liquid nitrogen.  The noise level will decrease
dramatically.  Take the terminator out of the N2 and let it warm up;
the noise level will return to normal. Be prepared to sacrifice the
coax you dunk into the N2; I've cracked a few rubber connector boots
this way.

>> And if that doesn't work, crawl up the spectrum a bit.  The higher in 
>> frequency you go, the more thermal noise you'll see.
>>
>Only up to a point. Past a certain point and the processes will start to
>roll off their energy production. 

It's a little more complicated than that. In general, as you go higher
in frequency the natural background noise power (lightning, sun noise,
galactic synchrotron radiation, 3K cosmic background, etc) decreases
rapidly, reaching a minimum in the 1-10 GHz range. Above that,
atmospheric components such as water vapor and oxygen again start to
contribute quite a bit of thermal noise.  (This low-noise window is
why the Search for Extraterrestrial Intelligence projects concentrate
on the 1-10Ghz range).  On the other hand, the noise contributed by
state-of-the-art receivers tends to increase with frequency, though
again the state of the art has gotten very good.

For cryptographic quality random numbers you want *only* local
receiver noise; if you rely on external sources like the galactic
background, your attacker also has access to them.

Phil