Fake Video Detection: A Deep Learning Based Approach

Fake Video Detection: A Deep Learning Based Approach

Sumit Tokhare1, Prof. Priyanka Choudhary2

Abstract: With increasing use of smart phones and social media applications, the sharing of videos has become extremely common. This has also led to the sharing of fake videos. Due to the size and complexity of the data being shared, it is almost infeasible for manual detection of video forgery. Since the data size to be analysed by time critical applications is enormous indeed, therefore the conventional techniques prove to be infeasible to detect fake videos with high level of accuracy, which primarily leads the focus to artificial intelligence and machine learning tools for the same. The proposed approach presents a frame decomposition, statistical feature extraction and machine learning based approach for fake video detection. The convolutional neural network has been used in this approach for detection of fake videos. The system achieves an accuracy of 98.5% with convergence at 1170 iterations. The results indicate that the proposed approach attains higher classification accuracy compared to existing techniques.

Keywords: Video Forgery, Artificial Intelligence, Machine Learning, Deep Learning, Convolutional Neural Networks (CNN), Classification Accuracy