Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site dciem.UUCP
Path: utzoo!dciem!mmt
From: mmt@dciem.UUCP (Martin Taylor)
Newsgroups: fa.human-nets
Subject: Re: HUMAN-NETS Digest   V7 #4
Message-ID: <610@dciem.UUCP>
Date: Sun, 8-Jan-84 16:48:45 EST
Article-I.D.: dciem.610
Posted: Sun Jan  8 16:48:45 1984
Date-Received: Sun, 8-Jan-84 19:24:10 EST
References: <763@brl-bmd.UUCP>
Organization: D.C.I.E.M., Toronto, Canada
Lines: 26

Several messages commented unfavourably on the idea of using voice
recognition for security. They are wrong.  Properly done, voice is
at present the most secure method of ensuring that the correct people
can get access to a physical facility, and presumably it could be
arranged similarly for a smart terminal to ensure access.  Texas
Instruments has for some time controlled access to its Corporate
Computer Centre by a voice security system (ask George Doddington
for details). In tests for the US Air Force, voice was more accurate
than any other single method. Naturally, for even better security
you mix methods.

Security methods can depend on the posession of an object or on
physical characteristics or on abilities.  A key or a card can be stolen,
fingerprints can be duplicated on a mould, but a voice response cannot
be duplicated when the person does not know what string the computer
is going to request.  If you have a security access card, the right
fingerprints AND the right voice, it is highly unlikely you are an
impostor. Incidentally, impressionists do only marginally better than
random "impostors" at getting past the voice check.

Voice security checks are vastly different from voice identification
in criminal investigation.  That's a much tougher problem.
-- 

Martin Taylor
{allegra,linus,ihnp4,uw-beaver,floyd,ubc-vision}!utzoo!dciem!mmt