Detection of Deep-Morphed Deepfake Images to Make Robust Automatic Facial Recognition Systems
Alakananda Mitra, Saraju P. Mohanty, Peter Corcoran, Elias Kougianos
Abstract
Face Morphing has emerged as a pervasive attack of Facial Recognition Systems. The rapid growth of Generative Adversarial Networks takes it to a complete new level. Deepfake or deep neural network based face morphing, a.k.a deep-morph attack, presents a significant threat to Facial Recognition System. In this paper, we propose a novel Convolutional Neural Network based detection method of deep morphed deepfake images which is suitable for IoT environments in smart cities. A high accuracy of 94.83% has been achieved for the DeepfakeTIMIT HQ dataset. This lightweight and fast network is a natural choice for IoT environments.
Topics & Concepts
MorphingComputer scienceConvolutional neural networkArtificial intelligenceDeep learningDeep neural networksFace (sociological concept)Facial recognition systemPattern recognition (psychology)Generative adversarial networkArtificial neural networkMachine learningComputer visionSocial scienceSociologyFace recognition and analysisGenerative Adversarial Networks and Image SynthesisDigital Media Forensic Detection