A Comparative Analysis of Deep Fake Techniques
Shubham Chaudhary, Rashid Saifi, Nishita Chauhan, Ritu Agarwal · 2021 3rd International Conference on Advances in Computing, Communication Control and Networking (ICAC3N) · 2021
Recent developments and evolutions in machine learning algorithms have directed to the emergence of first-class altered pictures that create video frames and have a striking similarity to real-world pictures that create video frames. This can have a disastrous effect on how people perceive digitally available knowledge or facts. Recent developments in machine learning and camera function have allowed pictures that compose video frames and sounds to be convincingly altered. These deep-fake films differ from actual videos by replacing the soundtrack, the lip movement, or the location where the video was captured (background). Many relevant tools and automated systems for detecting such deep fake films have been developed. Researchers employ techniques such as pose estimation, facial artefacts, temporal pattern analysis, background comparison, eye blinking, and mesoscopic analysis. In a unique statistical research, we want to present a descriptive review of these traditional deep Fake detection approaches.