FOCAL: A Forgery Localization Framework Based on Video Coding Self-Consistency

Forgery operations on video contents are nowadays within the reach of anyone, thanks to availability powerful and user-friendly editing software. Integrity verification authentication videos represent a major interest in both journalism (e.g., fake news debunking) legal environments dealing with digital evidence court law). While several strategies different forensics traces have been proposed recent years, latest solutions aim at increasing accuracy by combining multiple detectors features. This paper presents forgery localization framework that verifies self-consistency coding between frames, fusing information derived from set independent feature descriptors. The extraction step is carried out means an explainable convolutional neural network architecture, specifically designed look for classify artifacts. overall was validated two typical scenarios: temporal spatial splicing. Experimental results show improvement state-of-the-art splicing also promising performance newly tackled case splicing, synthetic real-world videos.