MesoNet3: A Deepfakes Facial Video Detection Network Based on Object Behavior Analysis

Deepfake is the process of manipulating objects with images and video. As a result of the development of deep learning techniques such as GAN, Deepfake has become closer to the truth. Many researchers are based on discovering deep fakes that were created by traditional methods. These methods produce often generate artifacts that may be subtle to humans. This paper can detect deepfakes that are perfectly created. Through the use of the new MesoNet3 algorithm to analyze behavior, facial expressions, and the appearance of an object based on a dataset. This paper consists of two stages. The first stage is to build a new MesoNet3 network that is trained on a set of data using a deepfake dataset. The second stage is to test the videos through the extraction of the face. After that, enter it into MesoNet3 and discover whether it is fake or not. The new MesoNet3 algorithm has proven its ability and accuracy in detecting fake video, compared to the old Meso-4. The accuracy of the MesoNet3 in detecting and distinguishing fake and real videos is %99.54.