From: wcs@anchor.ho.att.com (bill.stewart@pleasantonca.ncr.com +1-510-484-6204)
Date: Thu, 3 Mar 94 16:40:56 PST
To: cypherpunks@toad.com
Subject: Re: Standard for Stenography?
Message-ID: <9403040039.AA14605@anchor.ho.att.com>
MIME-Version: 1.0
Content-Type: text/plain


Hal Finney writes:
> Bill Stewart, Norm Hardy, and other list members who have more experience
> and who have discussed these issues in the past will I think agree that the
> correct approach is to separate the function of the steganography program
> to be a simple and clean insertion, and to have other components be
> responsible for assuring that what is inserted is statistically indistin-
> guishable from what is replaced.

It's somewhat of a tradeoff, though, since you really *do* need to have 
the system be convenient enough to use and standardized enough that
everybody will use it.  My own programming approaches tend to solve
this through reasonably clean programs connected by shell scripts
or C frontends grossly infected with Creeping Featurism;
the faults of this widely-used approach are well-known (:-).  

The important decisions, in my opinion, are whether to have an
explicit stego program or something that appears to be more general-purpose,
and whether to make sure the cyphertext you're hiding looks random.
If you're going to have an a program that admits to doing stego,
the main risks in having it do a fancy job are detectability
and portability, and it sounds like Jef's handling that well.
And Xenon's ranting has helped encourage someone to release Stealth-PGP:-)
so that's good.

Carl Ellison's "tran" program takes an interesting approach for data
scrambling - it takes a simple checksum of the first N bytes of the data, 
which is order-invariant (I think it was a byte-wise XOR?)
and uses it as a random-number seed for scrambling blocks of data;
it's easy to reverse because the checksum is the same after scrambling.
(I forget if the scrambling is also a self-inverse or not, which lets
you use one program for both directions; wouldn't be too hard to do.)
That might be a clean approach if you're still looking for a satisfactory
scrambling method, though you could also do bitwise things instead of 
bytesized, since you have to split the text out into bits for stego anyway.

		Bill