Object based forgery detection and localization in videos

Abstract

Video forensics involving passive approach exploits the statistical patterns of the video sequence to measure the integrity of video and detect forgery if it is compromised. Digital videos are susceptible to tampering due to the availability of efficient video processing tools for malevolent purpose. Object based video tampering disturbs under lying natural pattern of the video sequence. Quantal parameters of the spatial moments of a frame in a video provides inherent clue when object based forgery happens. The scalar quantities representing global characteristics of the frames in a video have been exploited to detect suspected frames. Thresholding approach is adopted to distinguish forged frames and original frames in a video sequence. Experiments on the subset of benchmark dataset demonstrate the efficiency of our approach.