Detecting Deepfake Videos Through CNN-MLP Model in Media Forensics
Megha Kandari, Vikas Tripathi, Bhasker Pant, Ayan Sar, Tanupriya Choudhury, Tanupriya Choudhury · 2024
Deepfake videos have become a growing concern in media forensics due to their ability to manipulate information and deceive individuals. The proposed method utilises a deep learning-based CNN-MLP model to extract features from the input video frames and then classify them into real and Deepfake. The proposed system used Celeb-df, which is a publicly available deepfake dataset, and achieved a high accuracy of $81.25 \%$ in detecting deepfake videos. Comparative evaluations with an alternative algorithm demonstrate our model’s superior accuracy, reinforcing its efficacy in discerning manipulated content. The proposed method can be integrated into existing video analysis systems to enhance their security and accuracy.