ENHANCING FACE RECOGNITION SECURITY: A COMPREHENSIVE STUDY ON ANTI-SPOOFING TECHNIQUES AND A NOVEL APPROACH BASED ON MTCNN WITH EMOTION AND POSTURE DETECTION

Abstract

Face recognition is a prominent technology with numerous applications, ranging from security systems to personalized user experiences. However, questions concerning the security and dependability of facial recognition systems have been raised due to their susceptibility to spoofing attacks. This thesis presents a comprehensive study on face recognition, including a literature survey on different approaches, their accuracy, and the challenges associated with them. The thesis specifically focuses on spoofing attacks on Face recognition system, such as photo attacks, video replay attacks, and 3D face mask attacks, which can deceive face recognition systems. Various anti-spoofing techniques, including liveness detection, motion detection, and analysis of facial reflectance and texture properties, are explored to mitigate the risks associated with spoofing attacks. Furthermore, a novel anti-spoofing model based on the Multi-task Cascaded Convolutional Networks (MTCNN) architecture is proposed. This model leverages the capabilities of MTCNN for face detection and recognition, and incorporates real time emotion detection and posture detection to enhance the anti-spoofing capabilities of the system. The proposed model is tested and evaluated, and its performance is compared with various other PAD models, demonstrating its effectiveness in mitigating spoofing attacks. The work in this thesis advances face recognition security by shedding light on various strategies, highlighting difficulties, and putting forth a fresh anti-spoofing model. In order to improve the robustness and accuracy of face recognition systems in the face of spoofing attempts, the results emphasize the need of combining various cues and real-time detection procedures.