Survey of non-intrusive face spoof detection methods

Abstract

Biometrics are distinct physiological characteristics used to describe individuals. Compared to the traditional access control methods such as passwords and Person Identification Numbers (PIN) which can be forgotten and shared easily, biometrics are widely used in authentication systems. Even though the accuracy of face recognition systems is lower than that of the systems using fingerprint, iris, etc. as the acquisition devices of the latter evade the affine and photometric transformations, recognition systems with the face as a trait are widely used due to the contactless and non-intrusive nature of the acquisition device-camera. As the cameras are in-built in most of the handheld and portable devices such as mobile phones and laptops, the uncontrolled and/or unregulated immediacy of sharing the photographs via messaging services and uploading on social networks entices the attackers to create spoofs to deceive a face recognition system. Hence, it is necessary to incorporate a spoof detection algorithm in recognition systems before revealing the identity. This paper gives an overview of the steps involved in the face spoof detection process, the various databases available, the different measures to discern between live and spoof images, aligned with the perceived observance, the binary classifiers used, and the performance evaluation parameters revealed in the literature.