AN OPTIMIZED MACHINE LEARNING MODEL FOR DETECTING FACE SPOOFING IN BIOMETRIC SYSTEM

Abstract

Face biometric-based access control systems are becoming increasingly prevalent in daily life; however, they remain susceptible to spoofing attacks. Addressing the need for robust and reliable anti-spoofing methods is essential. While deep learning techniques have shown promise in computer vision applications, their effectiveness in face spoofing detection is often hampered by the vast number of parameters and the limited availability of training data. This paper proposes a highly accurate face spoof detection system leveraging multiple feature extraction methods and deep learning. The system processes input video by extracting frames based on content, followed by cropping the face region from each frame. From these cropped images, multiple features are extracted, including Histogram of Oriented Gradients (HoG), Local Binary Pattern (LBP), Center Symmetric LBP (CSLBP), and Gray Level Co-occurrence Matrix (GLCM). These features are then used to train a Convolutional Neural Network (CNN). The proposed system's performance was evaluated using the Replay-Attack database, demonstrating superior results in spoof detection compared to other state-of-the-art techniques.