Technology advancement has made multimedia content widely available and easy to process. These benefits also bring ease to unauthorized users who can duplicate and manipulate multimedia content, and redistribute it to a large audience. Unauthorized distribution of information has posed serious threats to government and commercial operations. Digital fingerprinting is an emerging technology to protect multimedia content from such illicit redistribution by uniquely marking every copy of the content distributed to each user. One of the most powerful attacks from adversaries is collusion attack where several different fingerprinted copies of the same content are combined together to attenuate or even remove the fingerprints. An ideal fingerprinting system should be able to resist such collusion attacks and also have low embedding and detection computational complexity, and require low transmission bandwidth.

To achieve aforementioned requirements, this thesis presents a joint coding and embedding framework by employing a code layer for efficient fingerprint construction and leveraging the embedding layer to achieve high collusion resistance. Based on this framework, we propose two new joint-coding-embedding techniques, namely, permuted subsegment embedding and group-based joint-coding-embedding fingerprinting. We show that the proposed fingerprinting framework provides an excellent balance between collusion resistance, efficient construction, and efficient detection. The proposed joint coding and embedding techniques allow us to model both coded and non-coded fingerprinting under the same theoretical model, which can be used to provide guidelines of choosing parameters.

Based on the proposed joint coding and embedding techniques, we then consider real-world applications, such as DVD movie mass distribution and cable TV, and develop practical algorithms to fingerprint video in such challenging practical settings as to accommodate more than ten million users and resist hundreds of users' collusion. Our studies show a high potential of joint coding and embedding to meet the needs of real-world large-scale fingerprinting applications. The popularity of the subscription based content services, such as cable TV, inspires us to study the content protection in such scenario where users have access to multiple contents and thus the colluders may pirate multiple movie signals. To address this issue, we exploit the temporal dimension and propose a dynamic fingerprinting scheme that adjusts the fingerprint design based on the detection results of previously pirated signals. We demonstrate the advantages of the proposed dynamic fingerprinting over conventional static fingerprinting. Other issues related to multimedia fingerprinting, such as fingerprinting via QIM embedding, are also discussed in this thesis.