Siamese Meets Diffusion for Morphing Attack Detection

Biometric systems, such as facial recognition, are increasingly employed for secure identity verification. However, they are susceptible to morphing attacks, wherein images of different individuals are subtly blended to deceive biometric systems. Detecting these morphed images is challenging due to the subtle nature of the modifications. This paper proposes a novel approach that leverages diffusion models and Siamese networks to enhance morph detection capabilities. Diffusion models enable the extraction of fine-grained details by recreating the image after adding noise, while the Siamese network evaluates these details to accurately identify morphed images. Our method emphasizes improved detection accuracy, demonstrating significant potential to strengthen biometric security systems against morphing attacks.