Digital Forensic Technology For Video Forgery Detection

Digital media plays an increasingly important role in people's work and life, especially the digital video. Transmissing information quickly and efficiently, digital video has become the mainstream way of disseminating information. On the other hand, with the development of video editing software, video tampering becomes easier and easier, which makes digital video's credibility being questioned. Especially when the videos were brought to court as evidences, we have to identify its authenticity. In this context, digital multimedia forensics technologies was emerged, as one of its important branch, digital video forgery detection is used to determine the the originality and authenticity of the videos, and it has become an important weapon to ensure the originality and authenticity of the videos.Digital video forgery includes inter-frame forgery and intra-frame forgery,this paper focuses on inter-frame forgery. For two typical video tampering, frame deletion(or insertion) forgery and frame copy forgery, corresponding detection algorithm is proposed respectively. The main work is as follows:1. By studying the existing algorithms, we conclude the general video forgery detection principle and procedure. We aslo explain two video forgery detection article in detail, simulate their algorithms and points out the shortcomings of the algorithms.2. For frame deletion(or insertion) forgery,we propose a detection algorithm based on SPAM features. Calculating the similarity between adjacent frames by SPAM feature and determining the location of the forgery by detecting whether similarity between adjacent frames is smaller than the threshold value. We aslo present a dynamic thresholding algorithm based on local Chebyshev's inequality, the threshold can be adjusted depending on the video content, and using the local Chebyshev's inequality can reduce false positives or false negatives as the video scene is different at different times, resulting in a great similarity deviation. Experiments show that the algorithm has a better detection performance compared with the same type of algorithms and it's time consumption is relatively modest.3. For frame copy forgery, we propose a detection algorithm based on TCS-LBP histogram. TCS-LBP is evolved from CS-LBP which is extented from spatial to temporal. TCS-LBP has a stronger robustness compared with other spatial features. We also propose a modified algorithm,which takes correction processing on the copy-pairs sequence preliminarily detected, increasing the precision and recall rates of the algorithm. Even a static scene or a high frame rate video, the algorithm can detect the frame replication sequence effectively.