Robust Framework for Antispoofing Face Recognition Mechanism for Mitigating Image Attacks

Abstract

Abstract:The challenge here is that the appearance of human face can change drastically due to various illumination conditions and there are also many camera-related factors that may influence the quality of images, which makes it hard to differentiate images from a live person from those from photos. Due to these, simply asking "what's in the image (e.g., human skin)" tends to be unreliable. Another strategy is to use various image processing techniques to extract features that highlight the difference between images from live human faces and those from photographs. Work on fraud detection capabilities for face is still limited and a substantial part of it is based on the flatness of the captured surface in front of the sensor during an attack. This is also true for approaches that examine the 3D nature of the face by employing additional devices, which is much more realistic now with the introduction of affordable consumer depth cameras. With the help of the advancements in 3D manufacturing technologies, easily attainable facial masks take the spoofing attacks one step further and introduce new challenges for counter measure studies. The lack of protection against biometric spoofing attacks is not exclusive to face biometrics [2]. While it is possible to spoof a face authentication system using make-up, plastic surgery or forged masks; photographs and videos are probably the most common threats. Moreover, due to the increasing popularity of social network websites, a great deal of multimedia content, specially videos and photographs, is available on the web that can be used to spoof a face authentication system. To mitigate vulnerabilities, effective countermeasures against face spoofing must to be deployed.