Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!psuvax1!burdvax!sdcrdcf!randvax!jim
From: jim@randvax.UUCP (Jim Gillogly)
Newsgroups: sci.math,sci.crypt
Subject: Re: Encryption of Hoffman coded text
Message-ID: <292@markle.randvax.UUCP>
Date: Thu, 13-Aug-87 13:56:13 EDT
Article-I.D.: markle.292
Posted: Thu Aug 13 13:56:13 1987
Date-Received: Sun, 16-Aug-87 03:44:56 EDT
References: <10489@linus.UUCP] <7876@mimsy.UUCP> <10604@linus.UUCP> <1227@hropus.UUCP> <1034@faline.bellcore.com> <851@klipper.cs.vu.nl>
Reply-To: jim@markle.UUCP (Jim Gillogly)
Organization: Banzai Institute
Lines: 29
Xref: mnetor sci.math:1884 sci.crypt:523

>In article <1034@faline.bellcore.com> karn@faline.bellcore.com (Phil R. Karn) writes:
>>...  it is well known that increasing the entropy ("randomness") of the
>>plaintext before encryption is an excellent way to thwart cryptanalysis.
>>I've heard it from someone that should know that "knowing when you've found
>>the answer" is indeed perhaps THE hardest part of cryptanalysis.

In article <851@klipper.cs.vu.nl> biep@cs.vu.nl (J. A. "Biep" Durieux) responds:
>That sounds as if true, but I don't think Huffman encoding will make that
>so difficult (that is, Huffman encoding on a letter-by-letter basis), since
>English is also rather easily recognisable (from random bits) by it's inter-
>character statistics, i.e. the relative frequencies of groups of letters.

I agree with Phil, assuming the decoding tree is not known.  Huffman
coding (letter-by-letter or not) results in variable-length bit strings,
so that looking at a random piece of output you can't tell where the
letters (or groups of letters) begin and end.  This gives considerably
more room for error.  It's not terribly easy to decrypt Huffman-coded
stuff (even before a superencryption) without the decoding tree.
	    ------
If you know or guess that a particular word appears at a certain place
in the output, you can try to use that to build up the decryption.  That's
how I'd try to attack it anyway.  But I wouldn't expect to have it done
before lunch.

	Jim Gillogly
-- 
	Jim Gillogly
	{hplabs, ihnp4}!sdcrdcf!randvax!jim
	jim@rand-unix.arpa