Deep Fake Detection using ResNext Convolutional Neural Network (CNN) combined with a Recurrent Neural Network (RNN)

The utilization of digital video alterations has been observable for an extended period thanks to skillful deployment of visual enhancements; however, the latest progressions in deep learning have led to a significant rise in the credibility of artificial material and its availability. These AI-generated media, commonly known as DeepFake (DF), have become increasingly prevalent. Although creating DeepFake content using artificially intelligent tools is a relatively straightforward task, the challenge lies in the detection of such manipulations. Training algorithms to effectively identify DeepFake (DF) content is a complex endeavor due to the intricacies involved. In our pursuit, we have made notable progress in the identification of DeepFake material through the utilization of Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) architectures. The framework we have developed involves the application of a CNN for the extraction of features on a frame-by-frame basis, which are subsequently utilized in the training of an RNN. Through this process, the RNN is able to acquire the ability to determine whether a video has been subjected to manipulation, demonstrating its proficiency in recognizing temporal anomalies introduced by tools used in the creation of DeepFake content. Our projected results will be verified using a large dataset of fabricated videos obtained from established datasets. Our goal is to illustrate the competitive effectiveness of our system in identifying DeepFake material, emphasizing the straightforwardness of our framework while achieving reliable outcomes in this arduous undertaking.