From: "Alex Flanagan"
References: <email@example.com> <3F4C55C1.114C@despam.autobahn.mb.ca>
Subject: Re: the FBI and 1 time pads
X-Newsreader: Microsoft Outlook Express 6.00.2800.1158
Date: Thu, 28 Aug 2003 18:41:52 GMT
NNTP-Posting-Date: Thu, 28 Aug 2003 12:41:52 MDT
Organization: Shaw Residential Internet
> Well, not so clear. Eg use the sound card-- amplified gives random bits.
Assuming you pick up white noise (not brown or pink) with no trace of any
radio signals then, yes, these could be nicely random. I don't know how
shielded rooms would effect this, but a regular computer in a regular room
couldn't rely on bits generated this way being completely random.
> Those are not pseudo random numbers. They are actually physical
> variables which are probably as random and anything. Ie, unpredictable.
Well, keyboarding is non-random enough to be used as a method of collecting
biometric data, and to see patterns in how and where on the keyboard people
are typing (http://www.securityfocus.com/archive/1/333463). As for packet
arrival, patterns do emerge when you take a long term view (seconds) as
opposed to a short term one (packet to packet). With both of these there is
variation between events, but definite patterns that the variations just end
up adding noise to.
Well, once you find patterns you can eliminate them. Then you are just left
with the noise from above. This gives you a bunch of data within which the
message lies. Now you have to consider the unicity distance of the language
the message is in (for English it is about eight characters). What this
means is that once you have eight characters that make sense, the chances
are that you have decrypted the message.