Are we psychologically or culturally prepared for biometric
technology that can enable iris, voice and fingerprint recognition?
Danny Bradbury reports
Biometrics is a growth industry but, in the past few years, it
has not grown as quickly as some people had hoped.
The idea of being recognised by a computer on the basis of your
physical characteristics makes some people very nervous, promoting
fears of Big Brother scenarios. It has even brought the religious
community out of the woodwork, citing Revelations 13: 16-18 and
making noises about the mark of the Beast - a sign allegedly worn
by minions of the Antichrist that would enable them to recognise
and trade with each other. However, these people also said barcodes
were the sign of the devil, and not just a handy pricing mechanism
used in supermarkets. There's no pleasing some people.
A devilishly slow market
If the Antichrist wanted to use a particular technology as a
means of identifying his disciples, he would need a fair amount of
investment capital. In spite of efforts by companies such as
Compaq, which has included fingerprint recognition pads as options
on some PCs, the biometric market has grown slowly in the past few
years. Part of the problem has been the variety of different
standards and technologies constituting the biometric market. As in
many markets, fragmentation at the start of the growth curve leads
to slower adoption in the short term.
In a presentation to the International Biometric Industry
Association (IBIA) at the end of September 1999, analysts at
investment bank the McLean Group cited a Lehman Brothers prediction
that the biometric market would reach $400m (£274m) by 2004. As
e-commerce security becomes an increasingly popular issue,
biometrics could be a key solution. But then again, perhaps not.
After all, the most recent high-profile e-commerce frauds have
involved the theft of credit card numbers from an e-commerce
company's server and had little or nothing to do with customer
identification.
Fingerprints
Unfortunately for futurists, the shape of things to come in the
biometric market is relatively uninspiring. Don't expect any
world-beating new biometric technologies in the next few years,
because most of them have already been explored, and in many cases
made commercially available. Rather, the focus lies in making them
faster, cheaper, and more accurate.
Fingerprint recognition has been one of the most dominant
biometric technologies in recent times, with numerous vendors
offering solutions. Until recently, fingerprint recognition devices
operated mainly on an optical basis, where an image of a
fingerprint was scanned directly from the finger. These days,
companies are moving away from optical chips towards touch
sensors.
Peter Hawkes, chairman of the Association for Biometrics in the
UK, explains that such chips are more able to recognise greasy,
scarred or worn fingerprints - the latter making it useful for
manual workers.
According to Anil Jain, faculty member of the Pattern
Recognition and Image Processing Lab at Michigan State University,
solid-state sensors could cost as little as $1 (66 pence) per unit
in five years, while optical sensors currently cost a few hundred
dollars. Optical units are bulky, so you can't easily build them
into consumer devices such as laptops, mobile phones or smartcards,
he warns. Nevertheless, Compaq offers an optical PC Card-based
fingerprint scanner in its notebook devices.
Voice verification is another area in which companies are making
rapid advancements, but much work needs to be done. Companies such
as Irish firm Buytel are selling voice verification services.
Spokesman Ken Pilkington claims that its services are 99% accurate,
but some customers may still find this figure too unreliable. Jain
believes that when working with wireless technologies, the accuracy
of many voice verification technologies could drop to 94%.
Other problems for voice verification include the
signal-to-noise ratio. When talking over the telephone, issues such
as noise, wind in the earpiece and general bad call quality will
affect the final accuracy of the system.
Facing facts
Recognising people by sound is one thing, but visual recognition
is quite another. Facial recognition is already a commercially
available technology, with products from companies such as
Visionics (www.faceit.com) and Keyware (www.keyware.com).
According to a white paper from biometric expert Julian
Ashbourn, creator of biometric reference Web site Avanti, companies
must adopt realistic expectations when applying facial recognition
technology, especially when identifying faces in a crowd.
An alternative to facial recognition is to use the iris - the
individual part of the eye - for authentication.
"There is so much data it's like a very long barcode," says
Hawkes, who adds that some people may find it awkward to use.
Companies such as BT have tested it, but there are questions as to
whether the technology will be suitably accurate in all conditions.
For example, low light situations or neon street lighting may
reflect on the iris and distort the readings.
Apart from the questionable field results, there is also an
inherent fixed cost associated with the sensors involved, says
Jain.
"Sensing the iris requires more complicated processing, and
there is a little bit more co-operation from the user involved,"
Jain says.
The use of hand recognition is a much more established
technology, according to Ashbourn, who says at least one of the
commercially available products uses three-dimensional imaging to
build up an accurate, unique geometric picture of individual hands.
The advantage of this technology is that it is fast and unlike
voice verification, requires little training. This makes it useful
for situations where a high throughput of candidates is required. A
factory time clock system is a good example.
Several applications for biometric technology are in the offing.
Early last year, Oxford Micro Devices showed its chip-based
fingerprint recognition technology as a solution for "smart guns"
that only fire when the right person pulls the trigger. Other
applications include the inclusion of fingerprint recognition
technology in smartcards as a means of protection.
Other proposed applications include border control, and the
reduction of drivers' license fraud by checking applicants'
biometric data against a database to ensure that they are not
issued multiple licences. Both of these latter applications are
tougher nuts to crack, because they involve speed implications.
They entail one-to-many identification processes, where biometric
data is used to query a database for a match, rather than
one-to-one technologies, where the system has already been told by
the end-user who they are meant to be.
Universal standards
The real driver for biometric applications will be the
introduction of a standard application programming interface (API)
for software developers, so that they can code applications to work
with multiple devices at once.
There are several groups currently working on universal APIs for
software developers. The BioAPI Consortium, formed in 1998,
released the first version of its BioAPI specification in March
last year after consolidating its efforts with I/O Software and the
HA-API group, which were working towards the same end. Then, in May
last year, I/O Software sold its BioAPI technology to Microsoft,
which is likely to include the technology as a core service within
its Windows operating system in the future.
These two efforts will probably be the most significant drivers
for the uptake of biometric applications, and when Microsoft
includes the BioAPI and associated technologies purchased from I/O
Software into Windows, this will hopefully open the floodgates for
a concerted industry effort, which has previously been conducted on
an ad hoc basis.
There are still important non-technical issues to consider in
relation to the potential of biometric technology. For one thing,
personal preference and the co-operation of the user are vital to
its success.
Religion plays an important role, and not just among the fervent
few who are worried about the Antichrist. Muslim women in purdah,
for example, have most of their faces covered, excluding them from
any face recognition application. Similarly, many consumers may be
opposed to putting their hands or fingers into a scanner used by
the public on hygiene grounds.
Multiple solutions
These problems are not insurmountable, however. Hawkes is a
strong advocate of combination solutions, where multiple biometric
tests are used in unison. However, this brings its own challenges,
because more complex calculations are needed to interpret the
results and avoid incorrect conclusions.
Other technologies have been ignored in this article, primarily
because they are either unfeasible or simply on the drawing board,
with no medium or long-term commercial viability. Retinal scans,
for example, while technically feasible today, are unlikely to make
it into any commercial solutions, simply because people do not want
a bright light shone directly into their eyeball.
"It's expensive, intrusive, and it didn't catch on," says
Richard Norton, executive director of the International Biometric
Industry Association, adding that for some reason, whenever he
mentions biometric technology, everyone immediately thinks,
misguidedly, of the retina. Other proposed biometric tests have
included ear lobe analysis, and body odour readings. Armpit
scanners, anyone?
Timeline
2001: Fingerprint scanning deployed on some higher-end
laptops from companies such as Compaq
2002-3: Biometric technology included in Windows under
the .net initiative. Smart guns operate using fingerprint
technology
2003: Wider range of biometrically-enabled software
appears
2004: Market for biometrics reaches $400m (£600m)
worldwide.
Daily life under the biometric microscope
Boris has had his mobile phone stolen, but it doesn't matter
because the handset is protected by a fingerprint scanner. He calls
the mobile phone company to report it stolen and the system
verifies his identity by asking him the normal security questions.
However, we all know that these questions are far from secure, so
the server application in the call centre also analyses his voice
to verify his identity.
It's almost time for Boris to leave work, but he has to type a
letter before he goes. The dreaded Microsoft paperclip appears to
offer him unsolicited advice, but Boris' frown tells the computer
that he does not need help, so the paperclip disappears. He is due
to meet a friend for dinner, but needs some cash. He goes to the
cash machine and looks into the lens, where the bank can verify his
identity by performing an iris scan.
Here's looking at you, kid
IBM has been putting biometrics to good use at its Almaden
research laboratory. The company is working on a project called
BlueEyes that would give almost human-like characteristics to
everyday appliances, enabling them to "see" users and sense the way
that they are feeling.
The project uses a range of sensing technologies to analyse
human feelings. These include iris tracking and facial recognition.
The system would tell when you made eye contact, and begin to
react, analysing your face to see whether you were frowning or
smiling.
The iris tracking technology would enable the system to know
which way you were looking, so that you could intuitively move the
cursor, fine-tuning it by tweaking a mouse. The mouse itself would
be able to detect your emotional state by analysing your pulse and
temperature. The project pulls all of these developments together
to track user interests by monitoring patterns of activity and
working out what a user is reading by tracking their gaze.