Signal processing techniques for multimedia information security

The digital representation of multimedia and the Internet allows for the unauthorized duplication, transmission, and wide distribution of copyrighted multimedia content in an effortless manner. Content providers are faced with the challenge of how to protect their electronic content. Fingerprinting and watermarking are two techniques that help identify content that are copied and distributed illegally. This thesis presents a novel algorithm for each of these two content protection techniques. In fingerprinting, a novel algorithm that model fingerprint using Gaussian mixtures is developed for both audio and video signals. Simulation studies are used to evaluate the effectiveness of the algorithm in generating fingerprints that show high discrimination among different fingerprints and at the same time invariant to different distortions of the same fingerprint. In the proposed watermarking scheme, linear chirps are used as watermark messages. The watermark is embedded and detected by spread-spectrum watermarking. At the receiver, a post processing tool represents the retrieved watermark in a time-frequency distribution and uses a line detection algorithm to detect the watermark. The robustness of the watermark is demonstrated by extracting the watermark after different image processing operations performed using a third party evaluation tool called checkmark.