Research Of Video Watermarking Resist To Collusion Attack

With the rapid development of digital multimedia and Internet technology, there are more and more digital multimedia for our entertainment, on the other hand, it open new menace and challenges for the protection of copyright. As a potential effective method to protect copyright of digital video, video watermarking becomes an interest issue in the field of multimedia security and attracts many researchers; they have developed lots of video watermarking algorithm. Because of the high relativity of video frames, attracters can easily get collection of video frames for collusion attack; collusion attack is effective for lots of current video watermarking algorithms. Furthermore, with the emergence of new attack methods, the risk caused by collusion attack is more and more serious for digital video watermarking. In this paper, the author makes some research in the field of video watermarking resist to collusion attack; the main research in this paper lists as follows:(1) Firstly, the definition of Type I and Type II collusion attacks and how they can be used to estimate the watermark or host respectively are described, the Statistical Invisibility for Collusion-Resistant Digital Video Watermarking developed by Karen Su and the Spatially Localized Image-Dependent Watermarking (SLIDE) based on this rule are introduced. Then, a method of attacking SLIDE is proposed by extending the bound of the Type I collusion attack; based on image hash, an improved SLIDE algorithm is designed which can resist the new attack method. At the same time, the impact of using image hash on statistical invisibility of watermarking and the security of the improved algorithm is analyzed. The simulating results demonstrate that the proposed attack method is effective and the improved SLIDE algorithm is resistant to the new attack method.(2) The motion estimation based Type II collusion attack is described, which is proposed by Udit Budhia and Deepa Kundu when they studied Steganalysis using Statistical Invisibility criterion. The influences of the new attack method on current algorithms are analyzed. Within the framework of SLIDE, a method to resist the new attack method is developed, which is based on a novel feature point extract algorithm. This new feature point extract algorithm is designed based on the analysis of the characteristic that the feature point should hold for resisting the motion estimation based Type II collusion attack and the technology of image segmentation. Experiment results show that our feature point extract algorithm is robust and the digital video watermarking algorithm based on it has better performance than SLIDE algorithm at the aspect of withstanding motion estimation based Type II collusion attack.