Deepfakes Detection Methods: A Literature Survey

Recently, a great amount of concern has been attracted to the phenomena of DeepFake, which has been created for capturing and reenacting faces in a video and swap a face with someone else’s face using neural networks. In Deepfake technology, a computer-generated fake video shows fictional contents as real things. Many unbelievable applications in this technology are starting to be explored. Currently, malicious usages of fake videos are gaining in the computerized world like fake news, celebrity pornographic videos, revenge porn, and financial frauds. So, celebrities, politicians and famous people are facing mostly to Deepfake detection problem. Human’s naked eyes are weak to directly identify the difference between real videos and Deepfake videos because they are quite realistic. So, bothered people require an automated computerized Deepfake detection tool. This paper reviews existing Deepfake detection methods using traditional methods and deep learning technologies. Further this paper discussing the limitations of current methods and availability of datasets in the society. According to the literature, there is not a highly accurate and automated detection method to identify Deepfakes. The unavailability of efficient Deepfake detection method is a big challenge to the world due to the ease of generating Deepfake videos and their rapid spread. However, there are many efforts to solve this phenomenon and deep learning related methods show remarkable performance than other methods.