Exposing Digital Video Forgery By Detecting Motion-compensated Edge Artifact Based On Mpeg-2

In the digital multimedia era, digital cameras with high-quality have been widely used, the network everywhere also provides many image resources and video sharing websites, making digital images and video increasingly important in daily life. With the prevalence of high-performance computers and editing softwares integrated kinds of digital multimedia processing algorithms, it is becoming increasingly easier to tamper with digital video. Meanwhile, applications of video surveillance equipment have made dramatic increase in the amount of video data. Therefore, how to ensure the integrity and authenticity of the video data is of great significance.Deleting frame is one of the common and simple forgery methods, so detecting frame-deletion will have broad application value. A new approach to detect frame-deletion is proposed here, based on MPEG-2 codec, using motion-compensated edge artifact(MCEA) as a feature. MCEA is a side effect of the blocking impairment and motion-compensated prediction, appearing in video codecs that use block-based motion-compensated prediction, and brings in the addition of new high frequency which is not part of the original image content. Since the predictive coding mechanism is introduced in the widely applied MPEG-2 video coding system, and video sequences exist in group of pictures(GOP) structure. Deleting frame will lead to the restructuring of the original GOP, with some changes of original frame coding type, resulting in the decreasing of the temporal correlation in the GOP. MCEA has a closed relationship with the temporal correlation, once frame-deletion forgery appears, its change can be utilized for detecting video forgery.On the one hand, our approach introduces the MCEA impact factor to measure the changes of temporal correlation, thereby to determine the existence of forergy and locate the break point. On the other hand, video forgery can be further determined with the assistance of derivation trend of the MCEA group curve, whose surge point corresponding to the break point. Experimental results show that our approach achieves promising effect in video forgery detection, its overall detection accuracy grows with the number of deleted frames, and it's robust to the lossy compression.