DeepFake Video Detection Using Machine Learning and Deep Learning Techniques

The current issue facing the community is determining the legitimacy of online content, including movies and pictures generated by machine learning, in light of the development of Generative adversarial networks (GAN) and other deep learning-based DeepFake approaches. There is an extraordinary chance that we may severe violations of fundamental human rights combined with an inevitable, fundamental shift in the way people interact in society. Evidence of misinformation and manipulation of news headlines, medical (dis)information, and invasions of privacy have already been demonstrated. The objective of this proposed project is to efficiently identify DeepFake photos using an online image database. The categorization of real from false photos using convolutional neural networks and data from a sizable internet database is the main topic of this study. Our comparison of three distinct convolutional neural networks was our goal. 1) DenseNet, 2) VGG Face and 3) personalized CNN structure. Future research will examine whether real and false pictures cluster independently using unsupervised clustering techniques or auto-encoders. It would also involve using CNN visualization techniques to give our models more interpretability and transparency.