Strengthening Facial Biometrics: An Innovative Approach to Liveness Detection for Anti-Spoofing

Facial biometrics provide a dependable alternative to conventional authentication methods, with liveness detection playing a key role in distinguishing real users from spoofing attempts using photos or videos. Although significant progress has been made through techniques like motion analysis, texture analysis, and image quality assessment, challenges remain in ensuring reliable predictions, particularly with variations in facial angles in videos. This study introduces a novel approach that combines Dlib for facial recognition with YOLOv8 for enhanced detection and liveness verification. The system is trained on well-known datasets (NUAA and MSU-MFSD) as well as a custom dataset, enabling it to detect various spoofing methods, including photo, cut-photo, replay, and mask attacks. It achieves an impressive prediction accuracy of 98.5%. This research seeks to improve the reliability and consistency of facial recognition systems in critical security contexts such as border control and exam monitoring.