Reasearch On Digital Video Forensics

With the development of digital multimedia and network technology, digital videos are playing a more important role in our daily life. The advent of sophisticated video editing tools makes tampering digital multimedia data become much easier. If the tampered videos are used as news media or court exhibits, they will have an adverse impact upon society and judicial fairness. As an effective technology to monitor digital video resource, the blind video forensics has been a hot issue in the field of multimedia information security techniques.This paper primarily analyzes and studies two key points in the blind video forensics, the source video identification and the tampered video detection, and presents an appropriate solution respectively.This paper firstly proposes a source video system identification algorithm by analyzing the differences in the rate control module and the motion prediction module between different video coding systems. Three kinds of features, namely, rate features, quantization factor features and motion vector features, are extracted from reconstructed video. A support vector machine multi-classifier is used to build an intelligent computing system for video source identification. The experiment shows this proposed algorithm can effectively identify the source of video streams. Due to the re-compression coding is an indispensable link in the video post-processing, the double video compression detection technique is an effective forensic tool to recover the processing history of digital video resources. In this paper, the differences in the distributions of quantized DCT coefficients between the single and double MPEG-2 compression are analyzed in depth, and a statistical model of DCT coefficients is designed to study the statistical regularity in the distribution of quantized DCT coefficients which can be used to identify doubly compressed videos. Two commercial digital video cameras and multimedia editing softwares are introduced to simulate the process of double MPEG-2 compression. A series of experiments show that the proposed schemes can effectively detect double MPEG-2 compression at different output bit-rates.