Deepfake Video Detection by Using Convolutional Gated Recurrent Unit

Abstract

Rapid development in deep learning is making it easier to create fake videos known as “deepfake” videos in which human faces are swapped. Since deepfake videos are difficult to recognize by human eyes, it becomes important to automatically detect these forgeries and prevent their abuse. In this paper, we propose a deep neural network model to detect deepfake videos using a convolutional neural network (CNN) to extract frame-level features. These features are then used to train a convolutional GRU that learns to distinguish between fake and real videos. Evaluation is performed on the recently released Celeb-DF(v2)datasets where a state-of-art AUC score was achieved.