I would bet that comparatively few businesses would be interested in
deploying such devices if one-on-one wetware supervision of the
authentication process was required. Asked to rate the desirable attributes
of such a system, I bet that "automation" beats out "increased accuracy"
every time. What is the failure rate of human on human authentication,
using credentials and face recognition? High enough to support the entire
cost of developing a new branch of technology when all of the cost benefits
must come from reducing that one metric?
Scott Miller
Unimin Corporation
=> my opinions do not necessarily reflect those of Unimin Corporation, and
I have the scars to prove it...
"Jay D. Dyson"
<jdyson@treach To: Security-Basics List
ery.net> <[EMAIL PROTECTED]>
cc:
05/29/02 05:41 Subject: Unsupervised biometric scanners
PM little more than expensive toy.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Courtesy of Dr. Vann Harl.
Unsupervised biometric scanners are found to be more of an expensive toy
than a serious security measure
Thanks to Markus G. Kuhn:
An even more fatal blow to off-the-shelf unsupervised biometric
identification products was given recently by three authors in an article
in the well-respected German computer magazine c't:
Lisa Thalheim, Jan Krissler, Peter-Michael Ziegler: Krperkontrolle --
Biometrische Zugangssicherungen auf die Probe gestellt. c't 11/2002,
Heise Verlag, ISSN 0724-8679, p 114-, 17 May 2002.
An online English translation is now available on
http://heise.de/ct/english/02/11/114/
The team tested:
- six products involving capacitive fingerprint scanners
(Biocentric Solutions, Cherry, Eutron, Siemens and Veridicom)
- two optical (Cherry, Identix) fingerprint scanners
- one thermal (IdentAlink FPS100U) fingerprint scanner
(Atmel FCD4B14 sensor)
- Authenticam by Panasonic
- an iris scanner that is currently being marketed in the USA
and is scheduled to enter the European market in the near future
- FaceVACS- Logon, a technical solution for recognizing faces
developed by the Dresdner Cognitec AG
The authors "were able, aided by comparatively simple means, to outwit all
the systems tested" and concluded that "the products in the versions made
available to us were more of the nature of toys than of serious security
measures" and that "business should not treat the security needs of its
customers quite so thoughtlessly".
It is worth stressing that none of the deception techniques used are
really applicable in a *supervised* two-factor application, for example
where a border control or social benefits officer watches someone using
the finger or iris scanner in order to confirm the identity information
stored in a presented smartcard. The relevance of these attacks to the
discussion about the use of biometric features in a national identity
infrastructure is unfortunately sometimes misrepresented. I am still
convinced that both iris scanning and finger print recognition in a
*supervised* scan can be made easily several orders of magnitude more
reliable than human photo/face comparisons.
What currently marketed sensors lack is a really robust detection
technique for whether the detected signal comes from live human tissue,
and this still looks very much like an open research problem. Parts of
suitable solutions might be:
- tests of various involuntary reactions that require significant
effort to simulate, for example, is the iris pattern deforming
correctly when the pupils contract because of illumination?
- test whether the body part is functional, i.e. can the fingerprint
be detected from a finger that is typing fluently on a keyboard
or can the pupil inside the contracting iris read text at the same
time?
- is it possible to build low-cost spectrographic cameras/scanners that
can distinguish materials and tissues by using hundreds instead of
just three (red/green/blue) wavelength bands, etc.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (TreacherOS)
Comment: See http://www.treachery.net/~jdyson/ for current keys.
iD8DBQE89UsoGI2IHblM+8ERAsF8AKCCKkOSVBbHVATCaViwFxTMdMs4TQCeK9Md
nVbU3ylWoDsNb9T/E2zMsrs=
=n/ol
-----END PGP SIGNATURE-----
