A Hybrid Approach of Traditional Block-Based Deep Learning for Video Forgery Detection

Video forgery detection is a critical component of digital forensics and multimedia integrity verification. In an era where sophisticated video manipulation techniques, such as deep fakes and splicing, threaten the authenticity of visual content, the development of robust and efficient forgery detection methods is paramount. This research introduces a novel two-stage hybrid approach for video forgery detection, aiming to enhance accuracy and efficiency. The methodology integrates traditional block-based analysis with Convolutional Neural Networks (CNNs) to capitalize on local analysis and feature learning capabilities. The significance lies in addressing advanced forgery techniques and providing a comprehensive solution. The methods used combine meticulous spatial artifact examination with high-level feature learning, offering a versatile solution for video forgery detection. The hybrid approach achieved an accuracy of 79.31% and an F1-Score of 65.87%, significantly outperforming existing methods. This approach is robust to various types of video forgeries, such as face swapping, face reenactment, and splicing by providing a promising solution for video forgery detection that leverages the advantages of both block-based and deep learning techniques.