Perspectives on Information Hiding for Multimedia Security

当多媒体内容的安全性和质量得到保证时，一个完全数字化的多媒体丰富世界将成为现实。信息隐藏（水印、数字指纹等）在追求安全性方面越来越受欢迎。研究适合多媒体数据和网络的基于信息隐藏的安全技术是本课题的目标。本项目主要从以下不同的角度考虑提供多媒体安全的问题：(a)编码器/内容发起者的角度，(b)解码器/内容接收者的角度，(c)攻击者的角度，以及(d)网络的角度。从这些角度研究多媒体数据安全问题可以从端到端系统的角度更好地理解安全问题。以鲁棒性和嵌入能力为重点，研究基于感知的信息隐藏和基于信息论的信息隐藏与基于内容的水印之间的联系。提出了检测器依赖嵌入和博弈论公式来研究解码器与恶意攻击者之间的动态关系。在此基础上，提出了一种新的基于攻击的安全定义和计算信息隐藏容量的方法。最后，联合源信道编码和多重描述编码方法也可以用于鲁棒水印系统，可以处理网络损失。

A fully digitized multimedia enriched world will be a reality when security and quality of the multimedia content can be guaranteed. Information hiding (watermarking, digital fingerprinting, etc.) has been gaining popularity in this quest for security. Investigation of information hiding based security techniques that suit multimedia data and networks is the goal of this project. Primarily, this project considers problems in providing multimedia security from the following different perspectives: (a) Encoder/Content originator perspective, (b) Decoder/Content recipient perspective, (c) Attacker's perspective, and (d) Networking perspective. Studying the multimedia data security problem from these perspectives could lead to a better understanding of security issues from an end-to-end system view point. Connections between perceptually based and information theory based information hiding, and content based watermarking will be investigated with robustness and embedding capacity being the main focus. Detector dependent embedding and game theoretic formulation to study the dynamics between the decoder and a malicious attacker are proposed. A new attack dependent definition of security and methods to compute the information hiding capacity based on this definition are also proposed in this project. Finally, joint source-channel coding and multiple description coding approaches for robust watermark systems that can handle network losses are also of interest.