New Achievement: Still Liveness Detection
Facial recognition technology involves liveness detection and comparison-based recognition. With the increasing use of this technology in business, the security threat posed by face spoofing cannot be overlooked. Current anti-spoofing solutions include dynamic video face detection, eye blink detection, thermal infrared and visible light detection, etc.
Improving upon existing techniques, Matrix Space is developing an AI-powered liveness detection system which, with the help of 3D cameras, can accurately and consistently recognize a still human face.
Liveness detection is a crucial part of biometric authentication systems. It serves to foil spoof attacks by looking for physiological characteristics unique to living organisms.
Face liveness detection measures movement of the head, intake of breath, the red-eye effect, etc. Iris liveness detection measures iridodonesis, movement of eyelashes and eyelids, constriction and dilation of pupils, etc.
Still Liveness Detection
Still, liveness detection is one step further from traditional liveness detection technology. It does not require specific movements of the body in order to tell fake from real. For users, this new approach takes less time and can be useful in more situations, but at the same time, it makes a tougher demand on the system to guarantee accurate results.
In its liveness detection experiment, Matrix Space converted an RGB image to separate channels to generate individual LBP histograms before merging them into a histogram concatenation. Once this is done, an SVM decision can be made as to whether the image is real or fake.
Upgrading Still Liveness Detection
To help still liveness detection foil print, video, and 3D mask attacks, Matrix Space uses an infrared 3D camera in conjunction with AI to improve reliability when dealing with 3D subjects. With the help of a 3D camera, the still liveness detection system is able to achieve an accuracy of 94%.