Deep Learning-Based Generation And Detection Of Face Morphing Attacks For Secure Biometric Authentication

The rapid adoption of biometric authentication systems, particularly facial recognition technologies, has significantly improved identity verification in applications such as border control, digital identity management, and secure access systems. However, these systems remain vulnerable to sophisticated biometric attacks, among which face morphing attacks pose a serious security threat. In a morphing attack, facial images of two or more individuals are digitally combined to create a synthetic image that resembles multiple identities, allowing attackers to bypass biometric verification systems. Detecting such manipulated images is challenging due to variations in illumination, facial expressions, accessories, and image quality. This study proposes a robust deep learning–based framework for the generation and detection of face morphing attacks in biometric systems. The proposed approach integrates an advanced feature extraction mechanism with machine learning–based classification techniques to effectively distinguish between genuine and morphed facial images. To enhance detection performance, image preprocessing and enhancement techniques are incorporated to reduce noise and improve feature representation. Additionally, a diverse morph dataset containing both Morph-2 and Morph-3 images is utilized to simulate realistic morphing attack scenarios and improve model generalization across different facial characteristics. Multiple experimental evaluations are conducted using several publicly available facial image databases. The performance of the proposed model is assessed using evaluation metrics such as accuracy, precision, recall, F1-score, and detection error rates. Experimental results demonstrate that the proposed framework significantly improves morphing attack detection accuracy and provides a reliable defence mechanism for biometric authentication systems. By enhancing detection reliability and robustness, the proposed approach contributes to strengthening the security of modern facial recognition systems against identity fraud and biometric spoofing attacks.