Collaborative Research: Digital Fingerprinting: Information-Theoretic Analysis and Code Design

马里兰州大学帕克分校的Alexander BargU合作研究：数字指纹：信息理论分析和代码设计版权保护的指纹问题已经成为信息和编码理论研究的前沿，其动机是各种应用，如软件，图像，音频信号的数字指纹以及视频点播和相关应用的多媒体指纹。分发给系统用户的数据副本受到数字指纹的保护，即，为系统的每个用户单独设计的一种短的比特串，这样，单个用户在不被系统的分发者跟踪的情况下不能重新分发他们的内容。当一批用户合作创建数据的盗版副本，试图混淆他们的指纹时，追踪变得更加困难。本研究的主要目的是利用信息论和编码方法来确定一个具有可靠指纹数据的系统的最大用户数，并构造新的大规模指纹方案.本研究考虑了两种主要的指纹系统模型，一种是盗版者首先通过比较他们的数据副本来检测指纹的位置，然后对未注册的副本形成指纹，另一种是允许它们更改它们的数据文件，只要所创建的未注册副本与它们的原始副本相比满足某些失真约束。本研究将指纹识别系统视为固定或时变的多重存取通道，建立指纹识别在这两种模式下的容量限制。构造了新的指纹编码，并给出了有效的识别/跟踪算法和相关的性能界限。最后，本文研究了组合编码的性质，并设计了新的组合编码方案族，如分离和防帧码，具有识别父属性的码，以及可追溯码。

ABSTRACT0635271Alexander BargU of Maryland College ParkCollaborative Research: Digital Fingerprinting: Information Theoretic Analysisand Code DesignThe problem of fingerprinting for copyright protection has come to the forefront of research in informationand coding theory, motivated by varied applications such as digital fingerprinting of software, images, audiosignals as well as multimedia fingerprinting for video-on-demand and related applications. Copies of datadistributed to the users of the system are protected by digital fingerprints,i.e., short strings of bits designedindividually for each user of the system in such a way that individual users cannot redistribute their contentswithout being traced by the distributor of the system. The tracing becomes substantially more difficult whencoalitions of users cooperate to create a pirate copy of the data, attempting to obfuscate their fingerprints. Theprime objective of this research is to use information-theoretic and coding methods to establish the maximumnumber of users of a system with reliably fingerprinted data and to construct new, large-size fingerprintingschemes.This research considers two main models of the fingerprinting system, one in which the pirates first detectthe positions of the fingerprint by comparing their copies of the data, and then form a fingerprint for the unregistered copy, and the other in which they are allowed to alter their data files as long as the unregistered copy created satisfies some distortion constraints as compared to their original copies. Representing fingeprinting systems as multiple-access channels, either fixed or time-varying, this research establishes the capacity limits of fingerprinting under these two models. New fingerprinting codes are constructed equipped with efficientidentification/tracing algorithms and related performance bounds. Finally, this research investigates the propertiesand designs new families of combinatorial coding schemes such as separating and frameproof codes, codeswith the identifying parent property, and traceability codes.