Abstract

Face morphing attacks are an increasing concern in the digital world, posing threats to privacy and security. This study investigates the application of deep learning methods for generating and detecting face morphing attacks using convolutional neural networks (CNN) and autoencoders to model facial features realistically. A detection framework leveraging facial feature consistency analysis is developed to differentiate morphed images from real ones. The proposed system demonstrates high accuracy in identifying face morphing attacks, even when the morphed images closely resemble genuine ones. This study provides insights into deep learning applications for facial security and proposes directions for further research.