Performance and Security Improvements of Digital Watermarking Schemes and a Study of DRM Frameworks

• Main Authors: Chun-Hsiang Huang, 黃俊翔
• Other Authors: Ja-Ling Wu
• Format: Others
• Language: en_US
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/04404877508656046807

博士 === 國立臺灣大學 === 資訊工程學研究所 === 92 === Digital watermarking is regarded as a potential solution against the content piracy problem and is an important research topic in the past decade. After years of study, capabilities and limits of digital watermarking technologies have become more and more clear, and various applications of digital watermarking have been discovered. Designing an optimal watermarking scheme under certain fidelity/robustness constraints is still an open problem in the literature of digital watermarking. In this thesis, optimal watermark embedding is modeled as an optimization problem, and evolutionary computational skills are adopted to solve it. The fidelity and robustness requirements of digital watermarking are used as objective functions that guide the optimization procedure, respectively. As a result, a watermarking performance enhancement scheme is successfully proposed and evaluated. Furthermore, the proposed scheme possesses many desirable characteristics, such as asymmetric and independent embedding/extraction architecture that facilitates both optimal embedding and real-time extraction, blind watermark extraction, … etc. Visible watermarking is another protective mechanism in which visible but unobtrusive copyright patterns are embedded within contents that need to be protected. Attacking visible watermarking schemes means successfully recover embedded images or videos, under the requirement that only minimal user intervention is involved. In the proposed attacking procedures, copyright patterns consisting of thin lines are recovered by inpainting techniques. Furthermore, flat embedded areas are recovered by propagating color information from neighboring unmarked areas, and remaining areas are recovered by prediction using nearby recovery information. Acceptable recovered images can be obtained by the proposed attacking scheme. Suggestions that enhance the security of existing visible watermarking schemes are given, and simultaneously using both visible watermarking and fragile watermarking schemes are proposed as a solution against malicious attacks. In fact, the goal of content protection cannot be achieved with the watermarking technology alone. Digital rights management (DRM) – a combination of cryptographic techniques, digital watermarking schemes and rights specification languages – is a complete solution against content piracy. In this thesis, important DRM related standards, such as relevant parts within MPEG-21, CPPM/CPRM, CPSA, DTCP, HDCP, …etc, are surveyed and introduced. Furthermore, important commercial DRM systems, legislative measures and critical opinions are also included. In addition, a DRM framework protecting image distribution over digital networks and guaranteeing rights of both buyers and sellers are proposed and implemented. In this thesis, both optimal invisible watermark embedding and the security of visible watermarking schemes are discussed. Furthermore, a proprietary DRM framework is also implemented. Solving remaining open problems related to digital watermarking and DRM developments are future research directions of the author.