Detection of Deepfake Video Using Residual Neural Network and Long Short-Term Memory

Abstract

The appearance of web-based media has implied genuine and anecdotal stories introduced in such a comparative manner that it can now and then be hard to differentiate the two. Similarly, manipulation of real photos, audios or videos with the help of Artificial Intelligence techniques is done such that it is difficult to distinguish between the real and fake thus called Deepfake. It can happen to big celebrities, politicians, and to layman as well for some malicious purpose. Consequently, this procedure can end up being very threat to human culture subsequently expected to identify it appropriately. This paper intends to tackle this issue by proposing a model that uses Residual Neural Network (ResNet50) and Long Short-term Memory (LSTM) to detect video as fake or real. This approach tries to find flaws in the fake data left behind while its creation using neural based techniques like generative adversarial networks (GAN).