DEVELOPMENT OF AN IMPROVED GENERATIVE ADVERSARIAL NETWORK FOR DETECTION OF DEAPFAKE VIDEOS

Deepfake videos, which employ advanced machine learning techniques to create hyper-realistic fabricated content, have emerged as a significant challenge in today's digital landscape. The ability to detect these deceptive media is crucial for safeguarding against misinformation and preserving trust in visual content. In response to this pressing issue, this project presents the development of an improved Generative Adversarial Network (GAN) specifically designed for deepfake videos detection. The proposed enhanced GAN leverages cutting-edge advancements in the field of computer vision and adversarial learning to tackle the challenges associated with deepfakes detection. The research employs a novel architecture comprising multiple discriminator networks and a refined generator. The dataset includes a wide range of manipulation techniques and quality levels, enabling comprehensive testing of the GAN's robustness. Standard evaluation metrics, such as precision, recall, and F1 score, are utilized to quantify the detection performance and compare it with existing state-of-the-art deepfakes detection methods. The results demonstrate that the improved GAN achieves a significant enhancement in deepfakes detection accuracy compared to conventional approaches. The proposed model provides a efforts to combat the threat of deepfakes videos by presenting an advanced Generative Adversarial Network tailored for improved detection accuracy.