Tampering detection and localization in digital video using temporal difference between adjacent frames of actual and reconstructed video clip

The scientific, generalized and automatic methods for detecting forgery became the biggest challenge for scientists and researchers. This problem is true in case of all multimedia contents including audios, graphics and videos. It is harder when one doesn’t know the source and background of video in hand and still expected to establish authenticity of it. However, there are algorithms suggested which can work for such tampering in videos captured with static GOP structure. The problem becomes even more difficult when video is captured using adaptive GOP structure (AGS) scheme in which variable sizes of GOP structures are used to improve coding efficiency and to provide strong temporal scalability. In this paper, an algorithm is proposed which is a passive tampering detection algorithm based on comparison of temporal difference between adjacent video frames of actual video clip and its reconstructed version using intrinsic temporal fingerprints, which can work on videos captured using variable size GOP structures. Firstly, all the video frames are extracted from given video sequence. Then, temporal difference is calculated for each pair of adjacent frames in video’s actual and reconstructed from. Video is reconstructed using frame prediction error. Lastly, the calculated differences are used to find and localize tampering. Our proposed algorithm can effectively classify a video, irrespective of whether captured with fixed or AGSs, as genuine or forged using temporal difference between adjacent video frames in its actual and reconstructed form. Extensive experimental results show that the proposed method achieves promising accuracy in classifying genuine videos and forgeries. The results show that the proposed tampering detection algorithm can detect and precisely locate tampering with an average accuracy of 87.5%. © 2019, Bharati Vidyapeeth's Institute of Computer Applications and Management.