A Comparative Analysis of Deepfake Detection Techniques: A Review

Every person in this new generation has access to all of humanity's wisdom. Many technological possibilities are available. Nonetheless, we misuse this benefit by using deepfake for face swap. The deep fake method of machine learning, which is popular on social media, superimposes one person's face over another person's face. Deep learning, which has allowed researchers to produce deepfake images and videos considerably more rapidly and affordably, is the main ingredient in deepfakes. Even though the word “deepfakes” has a terrible image, more people are using the technology privately and practically. Although it is still fairly new, recent technological advancements have made it more difficult to tell the difference between deep fakes and synthetic images. Deepfake technology is evolving, and there is rising unease. The analysis of deep fakes has already been done by a number of researchers utilizing techniques like GAN, Convolution Neural Network with or without LSTM, Support Vector Machine, and Machine Learning. This paper analyzes the many techniques that have been used by researchers to identify Deepfake videos and rates the performance of multiple Deepfake video detection systems. According to our analysis, integrating CNN and the LSTM together produces superior outcomes and accuracy, which could also be improved by applying the concept of image enhancement.