Multimedia fingerprinting for multiuser forensics and security.

摘要

Recent development in multimedia and network technologies has made possible the ubiquitous sharing and distribution of multimedia over networks. However, illegal alteration and unauthorized copying of multimedia data pose serious threats to multimedia security and intellectual property rights, especially considering the ease manipulation of digital data. Therefore, it is critical to secure and protect multimedia content, and to ensure the integrity of rights by authorized users solely for intended purpose. Digital fingerprinting is an emerging technology to address post-delivery content protection and to enforce digital rights. In digital fingerprinting, unique identification information is embedded in each distributed copy, and is used to trace and identify the source of illicit copies. Such a traitor tracing is a fundamental problem in multimedia forensics, as well as an important tool for enforcing digital rights.; This thesis addresses various issues in multimedia fingerprinting. We first investigate the order statistics based nonlinear collusion attacks on digital fingerprinting, and analyze their effectiveness in defeating the fingerprinting systems. We also compare the performance of several commonly used detection statistics under collusion. We then examine the impact of scalable video coding and transmission on digital fingerprinting systems and collusion attacks. We analyze the effectiveness of the collusion attacks under the constraints that all colluders have equal probability of detection, and analyze the collusion resistance of scalable fingerprinting systems. We then consider the problem of traitors within traitors in digital fingerprinting, in which some selfish colluders wish to minimize their own risk of being captured while still profiting from the illegal redistribution of multimedia. We investigate the possible strategy by the selfish colluders to reduce probability of detection, and analyze their performance under the quality constraints. We also investigate the possible countermeasures by other colluders to protect their own interest. Finally, we investigate the secure distribution of fingerprinted copies for video streaming applications, and propose two secure fingerprint multicast schemes. We analyze their performance, including the communication cost and the robustness against collusion attacks, and discuss the tradeoff between the bandwidth efficiency and computation complexity.