Researchers develop new liveness detection method for facial verification
A group of academic researchers have developed a new challenge-response protocol for liveness detection, in which light from an LCD screen is reflected by a human face and captured by a camera to analyze the interval between the challenge and response.
Researchers from the Chinese University of Hong Kong, Fudan University, and Ohio State University published a paper on “Face Flashing: a Secure Liveness Detection Protocol based on Light Reflections” (PDF), which describes a light pattern emitted by the screen being reflected by a user’s face in a way unique to human faces. The reflections are captured by the camera, and the images applied to server-side liveness detection module prior to the actual face recognition module.
The researchers say the method was 97.3 percent to 98.8 percent accurate in different testing scenarios.
“Face Flashing” leverages the speed of light reflection, as well as the asynchronous refresh patterns of LCD screens, which form images from top to bottom, and cameras, which capture images from left to right. The timing requirements posed by this method make it virtually impossible to defeat both the liveness detection and the facial recognition simultaneously, according to the paper, particularly for dynamic 2D attacks, which the researchers say are the most common type.
The research paper was presented at the NDSS Symposium in February, Bleeping Computer reports.
BioID recently introduced a new liveness detection method for facial recognition on its BioID Web Service solution, and Brivas discussed its relative movement method of liveness detection with Biometric Update in January.