A Novel Texture based Approach for Facial Liveness Detection and Authentication using Deep Learning Classifier

Abstract

In today's digital age, face authentication stands as a pivotal method for secure user verification, offering convenience and heightened security. Our approach addresses critical challenges like low illumination, pose variation, and spoofing attacks by integrating advanced facial feature extraction and liveness detection with deep learning classifiers. Texture based facial feature extraction technique is proposed by combining feature-level fusion of Global (Gabor Wavelets) and Local (Local Binary Patterns) features, termed as GW-LBP. Moreover, the proposed texture based approach is also utilized for liveliness detection to analyze temporal and spatial variations indicative that the facial image belongs to live face or photograph or video (spoof). Using Our Database of Faces (ORL) dataset, this approach is evaluated using three deep learning classifiers: Convolutional Neural Network, ResNet50 and Vision Transformers which achieved an accuracy of 96.5%, 97.2% and 97.9% respectively. Moreover, the proposed approach demonstrates significant improvements in several other performance measures and feature extraction techniques and surpasses current cutting-edge methods as a resilience solution for user authentication.