From: Bill Stewart <stewarts@ix.netcom.com>
Date: Wed, 6 Mar 1996 13:05:47 +0800
To: cypherpunks@toad.com
Subject: Re: Is there any work on entropy-lowering schemes?
Message-ID: <199603040206.SAA07584@ix14.ix.netcom.com>
MIME-Version: 1.0
Content-Type: text/plain


At 07:46 PM 3/1/96 -0500, Mutant Rob <wlkngowl@unix.asb.com> wrote:
>I'm wondering if anyone has done any work on schemes to lower the
>entropy of a given stream.  Why?  Save you've got message M encrypted
>with a good cipher, but you're worried that it can be detected because
>even with stego, the entropy is a lot higher than normal 'random' data
>flowing through a network.

Peter Wayner's work on Mimic Functions does just this sort of thing.
You can describe a grammar, feed it random bits, and generate output that
has the right statistics and can be reversed to get the original bits.
His paper was on cs.cornell.edu a few years ago; don't know where
to find it now.  AltaVista yields a reference to the paper in Cryptologia,
and the Cyphernomicon has the following:
              - "They encode a secret message inside a harmless looking
              ASCII text file.  This is one of the very few times
              the UNIX tools "lex" and "yacc" have been used in
              cryptography, as far as I know.   Peter Wayner, "Mimic
              Functions", CRYPTOLOGIA Volume 16, Number 3, pp. 193-214,
              July 1992.[Michael Johnson, sci.crypt, 1994-09-05]

(When I read the Cryptologia reference on my browser, I don't get the
ligature in the middle of "Huffman coding"; YMMV. :-)