Preventing Spoof Attacks and Security Breaches
Thanks to the iPhone X, facial recognition—the ability to detect someone’s identity based on facial characteristics—has become widespread.
At MSU, Xiaoming Liu, assistant professor of computer science and engineering, is looking for ways to prevent recreating people’s faces in order to assume their identity or access private information. It’s called face anti-spoofing.
“Face anti-spoofing aims to classify between real human faces and spoof attacks, such as a printed paper with faces, a cellphone playing face videos and a 3-D mask, based on the image or video input,” Liu says.
A big part of the biometrics work being done at MSU focuses on ways to intercept spoofing or recreation methods by those intending to use information nefariously.
To prevent a spoof, researchers must test various methods against their technology. Liu’s lab is doing just that with a new framework, or algorithm, they’ve created.
“We recently proposed a novel framework to face anti-spoofing that leverages auxiliary supervision,” he says, referring to additional ways to test against face spoofing, including depth of faces and how the nose protrudes out farther than cheeks. “Our framework is the best of its kind out there.”
In addition to facial recognition, there are many ongoing research projects on related computer vision problems, one of which is called monocular pedestrian detection.
Liu and Ph.D. student Garrick Brazil have developed technology that can detect when a person is present on or close to the road—such as in a bike lane or on the sidewalk. The hope is that this technology can one day be used in autonomous vehicles so a simple camera could be equipped to determine if a person is present.
“Since last year, for this capability, we are actually the best among competitors on the benchmark evaluation, including Microsoft, the University of California, San Diego and Facebook,” Liu says.
Biometrics is becoming an increasingly common, and critical, component of our daily lives, and some of the nation’s foremost biometrics researchers at Michigan State University are leading exploration into applications that could make lives better by detecting health issues, preventing identity theft and enhancing security.
- Via MSU Today