A Study And Extended Application Research On Self-embeded Watermarking With Unifying Authentication Code And Recovery Code

With the rapid development of computer network and multimedia techniques, digital multimedia applications have been widely used in many areas. Given that multimedia information can be easily intercepted, copied, tampered, forged or deleted during their dissemination and utilization, which could damage their copyright, integrity, creditability and availability, the security issues about multimedia have become increasingly prominent and realistic. Traditional cryptography and digital signature technologies have some limitations in protecting multimedia information. The new digital watermarking technique, i.e. self-embedded watermarking, not only authenticates the image integrity, but also distinguishes the malicious alters from the legal manipulations, and locates the malicious tampered region as well as recovers the tampered region, providing an effective solution for image authentication. It has become a hot topic in information security in recent years.Multimedia mainly consists of texts, graph, images, audio, videos etc. This dissertation focuses on self-embedded watermarking on images. We analyze the conflicts of those existing self-embedded watermarking algorithms, and research into the similarity and difference between self-embedded watermarking algorithms and multimedia error concealment algorithms based on information hiding deeply. The main innovations are as follows:(1) Most existing self-embedded watermarking algorithms have the same feature that the watermarking signal embedded in a host image is divided into two parts independently:one part is used for authentication, and another is used for recovery. They are obtained and employed separately. The conclusion is that the conflicts among the elements related to evaluate the performance have become more distinct in those authentication-recovery independent algorithms, especially the conflict between capacity and robustness, capacity and imperceptibility, capacity and recovery quality, capacity and security, etc. So it is necessary to unify the authentication information and the recovery information.(2) Aiming at unifying the authentication information and the recovery information to reduce the pressure of the watermark capacity, a novel self-embedded watermarking scheme based on relationship function of corresponding inter-blocks DCT coefficient is proposed. The algorithm regards the function mapping result of DCT coefficients of logical neighboring blocks as watermark to unify the authentication information and the recovery information. The scheme can distinguish the malicious alters from the normal manipulations and recover the tampered blocks effectively. The algorithm extends the conventional two-value watermark by using real numbers as watermark signal, and provides a new way to express watermark in various watermarking schemes.(3) Most existing self-embedded watermarking schemes have no effective means to estimate their security performance, and this performance mainly obtains through experimental statistic. To settle this problem and unify the authentication information and recovery information, another self-embedded watermarking scheme based on characteristic clustering of image is proposed. The algorithm firstly clusters DCT direct coefficients and low-frequency coefficients which represent the characteristics of image blocks, and then embeds the encrypted sequence of clustering by introducing block index as watermarks into both middle-frequency domain of current image block and its successor block. Further, it verifies the blocks, which have been maliciously operated or tampered by comparing the consistency among clustering numbers and can be recovered using the clustering centers. The experimental results and theoretical analysis both prove that the scheme has superiorities over those self-embedded schemes based on DCT domain on the recovery effect of tampered area and the ability to resist various attacks while its security performance can be evaluated explicitly.(4) Due to most existing self-embedded watermarking algorithms based on blocks have weak ability to resist various geometric attacks, an adjuvant watermarking algorithm is proposed in spatial domain to estimate the type of geometric attacks as well as to eliminate the influence of those geometric attacks. Furthermore, the ability to detect various attacks including geometric attacks has been enhanced by combining the main watermarking algorithm and the adjuvant watermarking algorithm.(5) Most existing concealing algorithms cannot detect transmitting errors. After the similarity and difference between self-embedded watermarking algorithm and multimedia error concealment algorithm based on information hiding have been deeply explored, two error detecting and concealing algorithms based on characteristic clustering are proposed in spatial domain and transform domain respectively. It means that we have extended the self-embedded watermarking scheme introduced in chapter 4 to multimedia error concealing successfully. The proposed algorithms are different from the error concealing schemes based on contents estimate as well as from the schemes which directly embed various low resolution versions as hiding information into host images for error concealment. The experimental results and theoretical analysis both show that the proposed algorithms have the ability to detect errors but have less influence on stegano images and can get a comparable quality on tamper recovery, providing an effective way to protect image transmission in noisy channels.The theoretical analysis and experimental results in this dissertation both prove that the idea unifying the authentication information and recovery information in self-embedded watermarking algorithm is reasonable and effective. The idea has certain guiding meaning for self-embedded watermarking design as well as providing a new method to conduct the algorithms based on information hiding for multimedia error concealment.