Published Paper Details:

Face morphing attacks pose a significant threat to identity verification systems by seamlessly blending multiple identities into a single facial image. In this paper, we propose a deep learning-based approach for detecting morphed images using a Convolutional Neural Network (CNN) designed to identify GAN-generated artifacts and forensic inconsistencies. Our system dynamically loads or generates a detection model, ensuring adaptability to new morphing techniques. Additionally, we implement image forensics techniques, including noise analysis, edge consistency, JPEG compression artifacts, and hue variation detection, to enhance classification accuracy. A user-friendly graphical interface, developed using Tkinter, provides real-time feedback on the authenticity of an image by displaying real and morphed scores along with detailed forensic analysis. Experimental results demonstrate the effectiveness of our approach in distinguishing real and morphed images, making it a robust solution for face morphing detection in security applications.