Performance comparison of CNN models for detecting GAN generated deepfake images

Abstract

Deep fake images are content created by humans using the most advanced technologies. The technology imitates human expression to create fake images that are very similar to the real ones. Fake image creation has become very frequent due to rapid technological advancement. Again, with that advanced technology, it is possible to detect an image being fake or real. Therefore, we came up with the idea of giving a solution that which would be the most suitable network to rely on and why. Also which approach would be able to detect whether an image is real or fake most accurately. Using approaches of Neural Networks and Convolutional Neural Network, it becomes easier to decide and believe that which one's are real for sure. These deepfakes often creates misunderstandings among people. An image will be given to the system and it can verify whether this is a nely created fake image or not.