Detect Video Forgery by Performing Transfer Learning on Deep Neural Network

Abstract

Nowadays, the authenticity of digital image and videos becomes hard while the forgery techniques are more advanced. Given the recent progress on Generative Neural Network (GNN) development that may generate realistic images and videos, it becomes more difficult to detect the authenticity. In this paper, we expose a popular open source video forgery library called “DeepFaceLab” by making use of deep learning. We retrain the existing state-of-the-art image classification neural networks to capture the features from manipulated video frames. After passing various sets of forgery video frames through a well-trained neural network, a bottleneck layer is created for each image, this layer contains compact information for all images, and exposes the artifacts in forgery videos. We obtained above 99% accuracy when testing on DeepFake videos. In addition, we tested our method on FaceForensics dataset and achieved good detection accuracy.