Research On Multimedia Digital Signature Techniques

With the rapid growth of multimedia application technology and network technol-ogy, processing and distribution of digital multimedia products become much easier andfaster. Consequently, protection of intellectual property is becoming more prominent.However, multimedia has large volume of data and di?erent coding standards, and canbe represented equivalently in di?erent forms, and undergo various signal processingmanipulations like JPEG compression during distribution that may carry the same orsimilar perceptual content. This renders traditional cryptographic schemes, which canonly authenticate the binary representation of digital multimedia instead of its perceptualcontent, such as MD5 and SHA-1, inapplicable since they are bit-sensitive, i.e., even onebit change of the input will alter the output signature dramatically.Multimedia digital signature (also known as robust hash) is useful in protection ofintellectual property. It is perceptual features or short summaries of a multimedia object.The procedure of its extraction or generation is similar to conventional cryptographic hashfunctions: maps a multimedia object of arbitrary size to a short binary strings basedon the response of the human visual/auditory system (HVS/HAS) to the input signal.Consequently, perceptually similar objects yield similar signatures, while perceptuallydi?erent objects di?erent signatures. Hence, we call the procedure multimedia hashing.The technology of multimedia hashing can be applied to multimedia authenticationand identification, content indexation and analysis, digital watermarking, managementof large media database, copy detection, and so on. It is an emerging research area thatis receiving increased attention due to its wide prospective application.This dissertation focuses on some key problems related to multimedia hashing suchas general model and desired properties, unifying framework, applications and methods,especially the problems about the security, robustness and tamper sensitivity of imagehashing. The main works and e?orts accomplished by the author are described as follows:1. Firstly, since there still exists no perfect model up to now, this dissertationproposes a new one for multimedia digital signature, qualifies fully its desired propertiesand states its goals to describe the problem more clearly. Then, this dissertation presentsa unifying framework for multimedia digital signature, which can be applied to digitalimages, video and audio sequences, text documents, etc.Furthermore, this dissertation proposes a general method based on Principal Com-ponent Analysis (PCA) due to the independence on the low features extracted frommultimedia. The phenomena that PCA is the most e?ective with regard to energy con-centration and the base vectors (eigenvectors) vary from data to data, facilitate our...