|
|
Newsweek
'Very, Very Accurate'
A biometrics consultant argues in favor of
the technology used to fingerprint and photograph visitors
to the United States
January 7, 2004
By Brian Braiker
Visa-toting visitors to the United
States were greeted this week with a new kind of a welcome:
a demand for fingerprints and photographs. With little
fanfare, the U.S. Department of Homeland Security kicked off
its United States Visitor and Immigrant Status Indicator
Technology (US-VISIT) program to track the millions of
people who come to the country every year on business,
student and tourist visas--and to cull information in the
fight against terrorism. The program is a direct result of
federal legislation implemented in the aftermath of
September 11: The Patriot Act, the Enhanced Border and Visa
Entry Reform Act and the Aviation and Transportation
Security Act all mandated some kind of biometric identifier
to enhance public safety. The Visa Entry Reform, for
example, will not only require biometric identifiers on
every visa, but also on the passports of other countries
from which the United States doesn't require a visa.
Critics have voiced concerns
that the fingerprinting and facial-recognition procedures
will unduly delay travel and invade privacy while doing
little to track terrorists--especially those who, like most
of the September 11 hijackers, have no prior criminal
records. But Raj Nanavati of the high-tech International
Biometric Group research and consulting firm says the
technology being implemented at airports across the country,
if used properly, will not delay flights and over time will
prove a useful information-gathering tool. Nanavati recently
spoke with NEWSWEEK's Brian Braiker about biometrics, the
reliability of the technology, privacy concerns and what
U.S. travelers can expect when visiting other countries.
Excerpts:
NEWSWEEK: You're a consultant
in this field.
Raj Nanavati: As an
organization, we provide biometric consulting and technical
integration services as it pertains to the U.S.
border-crossing project and US-VISIT. Because we have been
working on behalf of the government, we cannot deploy the
system. So on this particular project we are working on
behalf of the government to help them evaluate the
technology.
How reliable is this
technology? Do these fingerprinting screens get smudged or
difficult to read over time?
The reliability and accuracy issue can be fairly complex and
encompasses a lot of different factors. At a high level, the
fingerprint device themselves are very accurate. There is
some degree of variance within the industry of different
products. The better fingerprint devices are using the
better algorithms [that] perform with error rates that are
substantially less than 1 percent. They're very, very
accurate in terms of making sure the correct person is who
they say they are and not falsely identifying people.
But people are operating this
technology, so there must be a component of human error.
In any kind of technology deployment there are
secondary issues. For example: do they clean the [screens
that read the fingerprints]; do people understand how to
place their fingers correctly. And all those vary from
application to application. Here you have a controlled
usage, you have someone supervising the person placing their
finger--they're not doing it at an unmanned kiosk. If the
[screen] is obviously dirty, it can be cleaned very simply.
For example, in New York City the people who get welfare
benefits are fingerprinted and between its uses, they just
wipe it off once. It takes half a second. No big deal. The
devices have been designed specifically to accommodate for
that, so they have certain materials on the surfaces
themselves that don't attract dirt and debris building up on
it and they work very, very well.
Do you know what the rate of
false positives is on this recognition technology?
For better or worse I can't quote you one number,
because I don't know that's it's been established exactly
what's going on with these fingerprints once they're
captured. And secondly, the rates can be modified. Given the
business requirements of the application, I can set the
false acceptance rate higher and the false rejection rate
might drop a little bit and vice versa. Or I could say, I
don't need as good an enrollment image because I want to
make sure I capture everyone's fingerprint knowing that I
might get more false matches later on, but everyone's
enrolled into the system. So there are all those factors and
then other quality checks and ancillary factors we talked
about--how do people place their fingers and those kinds of
things. But generally speaking, with the good quality
fingerprint scanners and with systems that are monitored and
used appropriately, you can get error rates below 1 percent
for false matches and false acceptance. Substantially less
than 1 percent.
What about facial recognition
technology?
Facial-recognition technology is not as accurate as
fingerprint. It works with standard photographs, so you
don't necessarily need new hardware that's customized for
the application. And in some people's eyes it has less of a
privacy concern because it's not possible to tie in with any
law enforcement databases. But as a counter to that, it's
not as accurate for reasons that are somewhat intuitive.
Your facial features change. Your fingerprint pattern
doesn't really change. But if you grow a beard or you
completely change your hairstyle or you're wearing glasses
or you gained 50 pounds or the camera doesn't capture at the
right angle ... If you place your finger correctly on the
fingerprint device and you've got a decent fingerprint, it's
probably going to capture a good print. With facial
recognition, if you're looking at an angle of 15 degrees,
the camera may not capture a good image, or there's a shadow
on half of your face. And those things complicate the
process.
Once they record your
fingerprints and take a picture of your face, what are these
devices searching for?
I'm not sure myself, and I'm not sure they've
released it. Based upon what I've read, they're comparing it
against a watch list of nondesirables. Whether that watch
list specifically is terrorists or felons or whatever the
case may be, I'm not certain. By way of example, the FBI has
a database of 40-plus million images in it. If the police go
to a crime scene and lift images, they can set it against
the FBI database. It's not instantaneous, but it comes back
within a relatively short time frame [allowing them to say]
"OK, there is a match or it isn't a match against
people in the database." So there is the ability of
fingerprint technology to work in very large databases, but
that factors into the time and the cost of the system. If it
were to take, for example, even five minutes to respond,
then it would slow down travel. So they're not matching it
to the full 40 million-person FBI database, but maybe they
have a database of felons--the top most-wanted, people who
have outstanding warrants and terrorists they've captured in
the past--and maybe it's a few thousand that are in that
database and they're matching that quickly.
How vulnerable is the system to
crashing or getting hacked into?
In terms of the technology, it's very simple to draw
an analysis to any IT infrastructure out there. There's
nothing unique with fingerprint technology or the camera
face-recognition system in terms of how technology operates.
If there's a power failure and they don't have backup, it's
not going to work. But the same thing would have to happen
at the reservation desk or the ticketing counter at an
airport. It's no more or less likely to fail than those are.
In terms of being able to hack into it, a similar rationale
applies there also. I would imagine that they deployed
appropriate standard IT protocols and firewalls and use
encryption and make sure there isn't the raw images being
processed through the Internet and things like that. The
information is secured in a way that it is perfectly
reasonable to be sending biometric data through their
network. When you communicate with your bank, you are
sending your account number and your password and all that
stuff is going through the Internet and people seem to be
very comfortable with that because precautions are in place.
The same things are happening here.
Brazil has bristled at this
policy and is retaliating by implementing their own
fingerprint system.
I think the idea or reciprocity is understandable and
acceptable. If we're requiring people from non-visa-waiver
countries to be fingerprinted when they come into the United
States, then requiring that U.S. citizens are to be
fingerprinted when they come through their country in my
eyes seems reasonable. They're not going through any unduly
invasive process--you're placing your finger on a piece of
glass for a second. Who cares? I think people have a
knee-jerk reaction to this technology and perhaps may not be
as familiar with exactly how it operates and the motivation
and impetus for utilizing that system. We're not doing it to
make life more difficult for Brazilians. We're doing because
there is a very clear and understandable security concern
that everyone in the world recognizes.
Is the technology they will be
using in Brazil comparable to and on the same level as the
technology used in airports here?
There are a variety of different fingerprint
technologies. We have consultants for perhaps a dozen
different countries who are considering using fingerprint or
iris or facial recognition also, and generally they would
operate in a similar fashion. Maybe instead of doing two
fingers it would be four fingers, maybe it would be one
finger. They general overall process would be somewhat
similar, I imagine.
As part of the
facial-recognition process, Muslim women are being asked to
remove their veils in public. Is there a fuzzy line between
respecting people's privacy, religious or otherwise, and
national security?
There is a balancing that has to go on between
security and privacy concerns. I think the officials that
are deploying the system took a careful look at specifically
the issue of veils because obviously that touches upon parts
of the world that have been associated with terrorism. They
probably looked at the factor and said, "Well, they
simply have to remove the veil for a few moments in order to
be verified." Practically speaking, when the official
looks at your passport, he needs to verify you manually that
you're the right person holding that passport. You need to
look at that person's face--there's no other way of doing
it. It applies to everybody.
What will airports look like in
20 years time?
I still see there being a variety of technology
because of different cultural and political attitudes in
different countries, if you're talking in a global sense.
For example in the United States, people get fingerprinted
when they take the bar exam, if you work in the securities
industry, if you're a teacher, if you're a hairdresser in
some states you have to be fingerprinted. In Australia,
you're essentially only fingerprinted if you're a criminal.
So using fingerprint technology by the government there at
airports for a civilian application is probably not going to
be acceptable. In fact, they're focusing more on facial
recognition. So different countries will pick different
technologies that are more in tune with their own cultural
values and norms. In terms of the process at any one given
airport, I think biometrics will be part of the long-term
permanent process as people become more familiar with the
technology as part of being authenticated at airports. It's
not a panacea and it's not designed for every single
deployment in the world. But at airports [where] you need
high security [and] where there's an infrastructure to
support the use of this technology, it is a very good idea.
|