BOSTON — One day you'll walk up to an automated teller machine and withdraw money without a card or pin number. Instead, the machine will recognize you automatically. Maybe it will match your face to your voice.
Such a machine may be closer than you think, as this week Diebold Inc. shows off an ATM that does just that. It is an indication of how far the field of biometrics - the statistical study of biology - has advanced.
Although investigated for years, biometrics is now being used by a few companies to create high-tech security products. Last year, for example, Diebold persuaded a South African bank to try out four ATMs that use fingerprint scanners to recognize customers.
With its current ATM demonstration, Diebold is using biometric technology from two smaller companies: face-recognition software developed by Visionics in Jersey City, N.J., and a voice-verification system from Keyware Technologies, which has offices in Woburn, Mass., and Brussels. When customers walk up to an automated teller, the ATM records their face - by recognizing its geometry - and asks them to say their special password or phrase into a telephone-like device. If both the face and voice match up, customers can make their transaction. Diebold believes the system is at least as secure as today's card and personal identification number.
The face-recognition software from Visionics is being tested in everything from computer access to international border crossings.
At one border crossing with Mexico, for example, the Immigration and Naturalization Service is already using Visionics's FaceIt system. The system automatically allows quick reentry for commuters who work in Mexico but live in the United States. All drivers have to do is make sure the video camera sees their face.
A Malaysian company is using FaceIt technology to create an airport security system that tracks passengers' baggage with an image of their face. Only when passengers actually enter the plane will the system allow their baggage to be loaded. The idea is to keep terrorists from checking bags in and not boarding the flight.
The system works this way: Hooked to a camera-equipped computer, FaceIt scans the video picture for faces. When it spots one (a technological feat, since computers typically don't distinguish visual images very well), the software tries to match the face to its database using statistical analysis. In other words, the software recognizes the geometry of the face (the angle of the cheekbones, etc.) and breaks it down into building blocks. If it then sees a face with the same building blocks put together in the same pattern, it approves the individual and gives access to whatever device it's guarding.
Since it uses only the geometry of the upper part of the face, it usually doesn't matter if someone shaves off a beard or gets a deep tan. The company claims that set at medium security, the system makes a mistake only once every 2,500 times, approving someone who shouldn't be allowed access. The security level can be set higher. But then the system works more slowly and is more likely to mistakenly reject people that it should let in.
"Information access is the most exciting area," says Joseph Atick, president and chief executive officer of Visionics. "You have a new world where computers are becoming the mechanism through which we connect to each other.... If there's an important e-mail, I want to put a face on it."
Individuals can buy a $200 version of the software, FaceIt PC, that protects files or a whole computer from prying eyes. Or you can download a free, 15-day evaluation of the program from the company's Web site (www.faceit.com).
* Send comments to firstname.lastname@example.org or visit my "In Cyberspace" forum at www.csmonitor.com