From: Sean Roach <roach_s@alph.swosu.edu>
Date: Mon, 20 Jan 1997 19:56:09 -0800 (PST)
To: cypherpunks@toad.com
Subject: Re: GSM crypto upgrade? (was Re: Newt's phone calls)
Message-ID: <199701210356.TAA27801@toad.com>
MIME-Version: 1.0
Content-Type: text/plain


At 11:41 PM 1/16/97 GMT, Adam Back wrote:
>
>Bill Frantz <frantz@netcom.com> writes:

>> As long as the man in the middle can't imitate a familiar voice,
>> this procedure is reasonably secure.
>
>This is the approach taken by PGPfone also.  If the value of the
>conversations was high (>$100,000?) passable voice imitation wouldn't
>be that hard I suspect.

I have long considered how easy it would be to use a sound card to modify
the human voice to match within certain tolerances the voice of another.

There are currently on the market, phones specifically designed to modify
the voice of the user so that kids can answer as adults, women can answer as
thier own protective boyfriends, bosses can answer anonymous calls as the
secretary, etc...

There are currently on the market keyboards that allow you to sample some
real world sound and use it as a voice in your music, (the model I saw, a
toy produced by Radio Shack, simply sped up or slowed down the sound to
achieve this.)

I have thought, if a machine were to take the incoming voice, analize
(apologies for spelling) it to get a spectrum signature, a pattern that can
be added or subtracted from another, and could then add the difference
between that and the victims signature to the users voice, then real-time,
on-the-fly con jobs would be easy.

The only thing that the user would be responsible for would be the accent,
and the day-to-day vocabulary of the victim.

I told a friend about this and he confirmed that such was available if you
knew where to look.