From: Eli Brandt <ebrandt@jarthur.Claremont.EDU>
Date: Tue, 8 Feb 94 13:07:09 PST
To: cypherpunks list <cypherpunks@toad.com>
Subject: Re: Clipper Side-step
In-Reply-To: <9402081756.AA28824@wavefront.wti.com>
Message-ID: <9402082106.AA16501@toad.com>
MIME-Version: 1.0
Content-Type: text/plain


> How about this as a way to stump Clipper?
[...]
> Encrypt data/message/information/recipe/whatever into the low-bits of
> the sound bite.

The low bits would probably be destroyed just by transmission over
your average voice line.  Worse, Clipperfones will compress the
input speech before encryption.  The only respectable audio
compression algorithms are lossy, and they will assuredly stomp on
your low bits.  Nor can you expect other modulations to survive
(e.g. the "data --> 212A --> Clipper --> 212A --> data" approach).

Given knowledge of the audio model used, you could take your data
stream and put it through the decompressor end.  The resultant
audio would be invariant under the lossy compression/decompression.

/--  sender  --\ /----          Clipper phone          ----\  /--  rcvr  --\
data->decompress->compress->encrypt,send,decrypt->decompress->compress->data
      \--   (cancel)   --/			  \--   (cancel)   --/

This would probably end up being manufacturer-specific and a real
pain.  Subverting a Capstone-based datacomm device would be easier.

   Eli   ebrandt@jarthur.claremont.edu