Deep Learning Techniques for Detecting Morphed Face Images: A Literature Review

Face morphing, as a digital image manipulation technique, is becoming increasingly prevalent, posing challenges in various fields, including biometric security, forensics, and identity verification systems. This paper provides a comprehensive review of the state-of-the-art of deep learning techniques employed in morphed image detection, including Convolutional Neural Networks (CNNs), Autoencoders, and Generative Adversarial Networks (GANs). These models have shown considerable potential in automating feature extraction, identifying subtle manipulation traces, and improving detection accuracy across different morphing techniques. However, as morphing techniques continue to evolve and produce more sophisticated attacks, current methods face challenges in detecting subtle artifacts, particularly in high-quality morphs. The review also highlights the issues related to dataset diversity and model generalization, noting that many approaches struggle with unseen morphing attacks and real-world scenarios, such as low-resolution, compressed, or printed & scanned images. Despite these challenges, deep learning techniques have made significant progress in improving Morph Attack Detection (MAD) systems. The paper suggests future directions for enhancing the robustness and generalization of these systems, focusing on expanding datasets, improving model adaptability, and developing methods that can detect increasingly sophisticated morphing attacks in real-world applications.