Video Inter-frame Forgery Identification by a Statistical Method

With the speed growth of digital technology, the recorded videos become a primary source of legal process in court. The video forgery identification becomes an important part of the legal process. This paper aims to propose an inter-frame forgery identification system using a statistical method. The system converts the extracted frames of the video into grayscale frame images. The system detects the edges in each frame using the Sobel edge detector. The system extracts effective feature vectors from the Hamming distance-based raw feature vectors by applying the Poisson Probability Mass Function (PMF). After that, the system identifies the given video that it is forged or not using the cross correlation and coefficient values of the adjacent frame. Video sources are gathered from UCF 101 data sets and SULFA data sets to test the proposed video forgery identification process. The experiment results show that the system can identify whether the video is forged or not in 69% of cases.