1994-03-08 - Re: more steganography talk

Header Data

From: Sergey Goldgaber <sergey@delbruck.pharm.sunysb.edu>
To: Peter Wayner <pcw@access.digex.net>
Message Hash: fc1288940faa73a9b7e34590efc12fb0202f4c78df2542157a638fe008f7469b
Message ID: <Pine.3.89.9403081330.A5604-0100000@delbruck.pharm.sunysb.edu>
Reply To: <199403051330.AA13596@access2.digex.net>
UTC Datetime: 1994-03-08 18:58:36 UTC
Raw Date: Tue, 8 Mar 94 10:58:36 PST

Raw message

From: Sergey Goldgaber <sergey@delbruck.pharm.sunysb.edu>
Date: Tue, 8 Mar 94 10:58:36 PST
To: Peter Wayner <pcw@access.digex.net>
Subject: Re: more steganography talk
In-Reply-To: <199403051330.AA13596@access2.digex.net>
Message-ID: <Pine.3.89.9403081330.A5604-0100000@delbruck.pharm.sunysb.edu>
MIME-Version: 1.0
Content-Type: text/plain

On Sat, 5 Mar 1994, Peter Wayner wrote:

> Sorry to be so distracted. This is a very interesting topic for
> me, but I've been bogged down with more prosaic topics. I think
> the Mimic FUnction implementation that I did is a very general
> standard for steganography. On the current level, it just deals
> with text, but you can make it do bits by just using the alphabet
> of just plain {0,1}.
> Here are the important points about it:
> 1) If the grammars are made complex enough, they can simulate
> anything you can compute with a computer. I.e. You can encode
> data in a Turing-complete way. 

I find it fascinating how complimentary cryptography and AI are!

> is done by translating RSA encryption into
> a context-free grammar.

I wonder if anyone has actually gone to all the trouble of 
developing some kind of binary CFG?  It should be easier to design 
than an equally effective human-language Turing-complete CFG.

> that there could be many practical "incomplete" attacks
> that worked in general cases. 

What kinds of "incomplete" attacks could possibly work against
Mimic functions implementing Turing-complete CFGs?

> 4) It is still unclear how to generate RSA-level strength
> with Mimic Functions. 

Can't you simply use a Turing-complete CFG, and meta-CFG?
Do such things exist on computer media?

> Understanding what makes grammars hard and easy to grok is a 
> hard question. 

Why not just ask an AI?  :)

> 6) The process is slightly difficult to implement, but I've
> got two running versions (as I've mentioned before on the 
> list). One in C and the other in Pascal.

Do you know if anyone has ported either of those over to anything
other than the Mac?

Good to have you join the discussion, BTW...