Gan-Enhanced Real-Time Detection of Deepfakes Videos
Amber Fatima, Pintu Kumar Ram · Journal of Artificial Intelligence and Capsule Networks · 2024
Artificial intelligence is used by the deepfake mechanism to create videos that are remarkably realistic yet fraudulent, seriously undermining the legitimacy of digital media. In this study, AI techniques are used to examine real-time deepfake detection, with a particular focus on Generative Adversarial Networks (GANs), Recurrent Neural Networks (RNNs), and Convolutional Neural Networks (CNNs). The DeepFake Detection Challenge (DFDC) dataset was used to test the models after essential features were extracted using a frame-based technique. According to the findings, GANs outperformed CNNs (83%), RNNs (85%), and other neural networks (ANNs) with an accuracy of 88%. With a 3% improvement over RNNs and a 6% improvement over CNNs in accuracy, as well as improved recall and precision measures, GANs proved to be superior generative feature learning. This proposed study demonstrates the potential of GAN-based techniques for reliable detection in difficult real-time.