ADFB: Anti-deepfake Framework for Facial Biometric Authentication Systems

Facial biometric authentication, which leverages distinctive facial features for identity verification, is a widely used biometric identification method. However, deepfakes synthetic images or videos generated through sophisticated deep learning techniques pose significant challenges to the reliability of these systems by mimicking real facial data. The presence of deepfake-generated data in facial biometric databases threatens their accuracy, potentially leading to incorrect identifications and fraudulent authentications. This research undertakes a comprehensive study and comparative analysis of AI-driven facial biometric techniques published between 2019 and 2024, including an extensive survey and evaluation of existing deepfake facial detection systems. The paper outlines the workflow of a proposed facial biometric authentication system designed to counter deepfake threats, ensuring secure and reliable identity verification. The model achieved an accuracy of 96.59% on a dataset of real and fake images for deepfake detection and an accuracy of 99.56% on a face images dataset for facial biometric authentication. These results demonstrate the effectiveness of the proposed framework in enhancing the security and reliability of facial biometric systems against deepfake threats.