Secure Facial Recognition Systems: A Machine Learning Review of Spoofing Detection via Parameter Quality Metrics

Abstract

Facial recognition systems are widely used for authentication and security, but they remain vulnerable to spoofing attacks using photos, videos, or 3D masks. This study proposes a machine learning-secured facial recognition system integrated with spoofing detection based on the quality of parameters to enhance reliability and security. The system leverages deep learning techniques for feature extraction and biometric authentication while incorporating image quality assessment metrics, such as illumination consistency, texture analysis, and liveliness detection, to differentiate genuine faces from spoof attempts. Additionally, a hybrid model combining Convolutional Neural Networks (CNNs) and transformer-based architectures is used to enhance classification accuracy. Experimental evaluations on standard benchmark datasets demonstrate high accuracy, robustness against adversarial attacks, and improved generalization to diverse spoofing techniques. The proposed approach significantly reduces false acceptance and rejection rates, ensuring secure and efficient facial recognition for real-world applications in access control, financial transactions, and identity verification systems.