|
|
|
Clarkson University Associate Professor of Electrical and Computer Engineering Stephanie Schuckers demonstrates how simple casts made from Play-Doh and gelatin can fool most fingerprintrecognition devices. |
These scenarios for beating high-tech security and identification systems may not be as farfetched as you think.
Biometrics is the science of using biological properties such as fingerprints, an iris scan, or voice recognition, to identify individuals. And in a world of increasing security measures, the field is rapidly expanding.
"Biometric systems automatically measure the unique physiological or behavioral "signature" to authenticate an individual's identity," says Stephanie C. Schuckers, associate professor of electrical and computer engineering at Clarkson University, Potsdam, N.Y. Biometric systems, she explains, are popping up in hospitals, banks, even college residence halls — to authorize or deny access to medical files, financial accounts, or restricted areas.
Biometric systems are prone to "spoofing" or attacks designed to defeat them. Spoofing is the process of overcoming a system through introduction of a fake sample. "Digits from cadavers, fake fingers molded from plastic, or even something as simple as Play-Doh or gelatin, can be misread as authentic," she explains. Schuckers' research is looking for effective safeguards.
With funding from the National Science Foundation, the Office of Homeland Security, and the Dept. of Defense, she is currently assessing vulnerability in fingerprint scanners and designing methods to address it. To assess how vulnerable fingerprint scanning devices are, Schuckers and her research team made casts from live fingers using dental materials and Play-Doh to create molds.
They also assembled a collection of cadaver fingers. The researchers tested more than 60 of the faked samples. The result was a 90% false verification rate. The machines could not distinguish between living and fake samples.
In live fingers, perspiration starts around the pores and spreads along the ridges creating a distinct signature. The research team designed a computer algorithm that detects this pattern. With the new system, less than 10% of the spoofed samples fooled the machine.