From: hfinney@shell.portal.com (Hal Finney)
Date: Fri, 8 Oct 93 14:05:47 PDT
To: honey@citi.umich.edu
Subject: Re:  second order homophonic substitution
Message-ID: <9310081626.AA17600@jobe.shell.portal.com.shell.portal.com>
MIME-Version: 1.0
Content-Type: text/plain


This homophonic cypher sounds interesting.  If the authorities demand
your keys, you could scramble each row of the matrix (scrambling each
row separately), so that only the dummy message can be recovered.

Real ascii messages have a character set of about 2^6, so the actual
size of the key matrix will be 2^12, and that means that each entry will
be about 12 bits.  For full generality in handling binary data the
character set would be 2^8, meaning a matrix of size 2^16 entries with
each entry being 16 bits.  This is a 128K byte key, which is pretty
cumbersome.  Also, the cyphertext is twice as big as the plaintext,
which will stand out too.

Plus, once the authorities see your decryption algorithm it may be
pretty obvious that it was designed for this specific purpose, and
whatever pressures they applied to make you reveal the key may now
be redoubled until you reveal the "real" key.

A one-time-pad has the advantage that the key is the same size as the
file, and there is no size expansion in encryption, plus it's a plausible
approach to use for high-security encryption.  It will take less space
and still allows for multiple decryption.

Hal
hfinney@shell.portal.com