Portal de Periódicos da CAPES

A Review on Deepfake Media Detection

2022; Springer International Publishing; Linguagem: Inglês

10.1007/978-981-19-2130-8_28

2367-3370

Rajneesh Rani, Tarun Kumar, Mukund Prasad Sah,

Generative Adversarial Networks and Image Synthesis

Deepfake is a machine learning and artificial intelligence-based technique which is used to generate fake faces and replaces them in a video or image. These days deepfakes are mostly used to spread fake news, audio, video, etc. Deepfakes are also used in political campaigns to manipulate the videos of leaders and spread hatred. Today, it has become very easy to generate deepfake images as there are various commercial softwares available for generating deepfakes; also, there are various free of cost apps available like faceapp, fakeapp, etc. The website thispersondoesnotexist.com generates a fake person image that does not exist every time you click it. The automation in video manipulation generates more threat to original content as it is becoming more and more easier to manipulate images and generate fake news. It can be very dangerous in the upcoming time to detect what is fake and what is real. The main component which makes deepfakes more and more real is general adversarial networks (GANs); by using GAN, we can generate high-quality deepfakes which cannot be detected by the human eye. There are various techniques generated to detect deepfakes, but to the best of our knowledge, we can say that there is no foolproof method to detect deepfakes, and there is a strong need for a technique which can prevent current facial recognition systems from deepfakes. In this paper, we try to give a brief review of existing deepfake detection techniques. There are many techniques developed in this area but most of them can be categorized in facial artifacts, neural networks, 3D head position.