1993-06-20 - Re: that cipher I wrote (pseudo random generators)

Header Data

From: gnu (John Gilmore)
To: cypherpunks@toad.com, gnu
Message Hash: 50517258f4dfe27c6d4a7c196f0baf7e23b7c797ae99543e17a351bac8c90783
Message ID: <9306200157.AA06857@toad.com>
Reply To: <9306190933.AA12046@hairball.ecst.csuchico.edu>
UTC Datetime: 1993-06-20 01:57:35 UTC
Raw Date: Sat, 19 Jun 93 18:57:35 PDT

Raw message

From: gnu (John Gilmore)
Date: Sat, 19 Jun 93 18:57:35 PDT
To: cypherpunks@toad.com, gnu
Subject: Re: that cipher I wrote (pseudo random generators)
In-Reply-To: <9306190933.AA12046@hairball.ecst.csuchico.edu>
Message-ID: <9306200157.AA06857@toad.com>
MIME-Version: 1.0
Content-Type: text/plain

> What characteristics of the multiplier and modulator provide large periods?

Don't bother with looking for a large period (though Knuth spends
about half a book on pseudorandom number generators).  The problem is
that the sequence is predictable.  Given ten sequential values from
anywhere in the sequence, I think there are algorithms that will
determine the sequence.  This is from dim recall of some Crypto '90
presentations.  Perhaps someone has the papers in front of them, or
can re-derive the results?

Given plaintext XOR'd with a sequence, you can make pretty good guesses at
ten values in the sequence, and if you have to try a few thousand guesses,
it will still only take minutes or hours to crack.