Face Liveness Detection to Overcome Spoofing Attacks in Face Recognition System

Abstract

Face recognition is widely used across several biometric applications. Hence, it is very important that the face recognition system is completely secure and is not prone to any threat. In recent research, it is observed that the face recognition system is vulnerable to many attacks such as spoofing, morphing or masking. These attacks pose a severe threat to the security of the face recognition system. In this paper, a novel method is proposed for face liveness detection to overcome spoofing attacks, and then, it is followed by face recognition. The proposed face liveness detection method is designed to identify whether the captured face image is of the live person or spoofed face of the person. Further, the face liveness detected image is processed for face recognition using OpenFace. In case, if face liveness detection fails, it is not further processed for face recognition. Thereby, the method overcomes the spoofing attacks on the face recognition system in the primary input step itself. The proposed system for face liveness detection is developed using convolution neural networks (CNN) for deep learning. The face liveness method has experimented on author’s own dataset and NUAA anti-spoofing dataset. The proposed face liveness detection method has been compared with state-of-the-art methods, and the comparative study shows that the proposed method has performed significantly better than state-of-the-art methods for face liveness detection. The face recognition step using OpenFace is experimented using author’s own face image dataset and benchmark dataset NCKU. The experimental results show that the proposed face liveness detection method has yielded an accuracy of 94.82%, and for face recognition, the accuracy obtained is 98.14%, in respect of own dataset.