Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!ll-xn!mit-eddie!uw-beaver!uw-june!uw-entropy!martin
From: martin@entropy.ms.washington.edu (Don Martin)
Newsgroups: sci.crypt
Subject: Re: One time pads?
Message-ID: <784@entropy.ms.washington.edu>
Date: 4 Mar 88 06:35:08 GMT
References: <4209@june.cs.washington.edu> <1988Feb15.151522.5094@utzoo.uucp> <591@gtx.com>
Organization: UW MathStat, Seattle
Lines: 25
Summary: Hash codes to make english random


There has been some discussion of the nonrandom nature of
english text in the context of one time pads. Two remarks:

1. A Beale cipher takes advantage of the key and the message
having similar frequency distributions. This cipher appears
to be quite safe if the key is as long or longer than the
message.

2. I did some expermentation with cryptographic hash functions
a while back. As a test, I hashed the unix spelling dictionary
which is about 26000 words into 3 decimal digit numbers and used
a chi-square test on the resulting frequencies. I did not test
for serial correlation type structures. It took more tinkering
with the hash function than I expected to get a good frequency
distribution. However, I was convinced that a hash function that
looked random to the usual statistical tests was possible and
that the function could be designed so that it was difficult to
find a partial inverse.
    This type of hash function is handy both as a hash function
for the usual cryptographic uses, such as signatures, and as a
good way of seeding random number generators. However, you must
use many more input characters than the function puts out to avoid
hidden redundencies.

Don Martin