From: ghio@cmu.edu (Matthew Ghio)
Date: Wed, 9 Aug 95 18:19:47 PDT
To: cypherpunks@toad.com
Subject: Re: this looked like it might be interesting
In-Reply-To: <199508091852.UAA23821@utopia.hacktic.nl>
Message-ID: <m0sgM6l-000vpPC@myriad>
MIME-Version: 1.0
Content-Type: text/plain


Hmm.. This is interesting indeed.  It is a 32-round cipher operating
on a 64-bit block.  It has an 80-bit key with a 16-bit cryptographic
checkword.  This seems a lot like a certain hardware-based encryption
system some TLA spooks were pushing about a year ago.

If this isn't Skipjack, someone sure went to a lot of trouble to make
it look like skipjack.  It's also possible that it is an early development
version of what later became Skipjack/Clipper.  It is dated February 1989
and July 1991, which would be consistent with NSA's claim that they had
been working on Skipjack for about 5 years.  It also mentions a "S-2"
revision.  It'd be interesting to try to see if it will interoperate with
a real clipper chip, but I wouldn't bet on it.

Hal Finney noticed that the coding style seems sloppy or amateurish.
This is probably just an attempt by the programmer to hide his tracks.
The RC4 stuff also had unnecessary operations in it, presumably for the
same reason.

Hal also comments on the use of XOR instead of bit permutations as in
DES, and the use of 8-bit table sizes, which would make a software
implementation easier.  This is odd, considering that the algorithm
purports to be designed for hardware.  However, NSA did say that part
of the clipper algorithm was in software, which was designed to be
erased if the chip was tampered with...