Research On Key Technologies Of Forensic And Trailing For Unauthorized Disseminating Of Digital Images

It is convenient to people because of the wide spread of digital images. But their unauthorized spread also brought a series of security issues in the open network. Therefore, how to solve them is one of the urgent issues of information security. Apparently, digital watermarking as a significant branch of information hiding is an effective means of the digital copyright protection and the integrity authentication. The thesis mainly makes researches on forensic and trailing for the unauthorized spread of digital images. The content can be summarized into three parts including robust watermarking, fragile watermarking, and multi-watermarking listed as follows.1. A robust watermarking algorithm based on image edges is proposed for a balance between the robustness and the invisibility of watermark. It uses Prewitt operator to detect the edge of cover image, select the edge points and embed watermark into the gradient directions of the edge. Due to not only the masking effect of human vision at the edges, but also the importance and stability in image processing, the algorithm has a good performance on the robustness and invisibility. The detailed theoretical analysis about resisting attacks is given in the thesis and verified in experiments. In addition, combination with the masking effect of human vision, we propose an evaluation method of the watermarked image quality, which is verified in the experiments and shows the good invisibility of the proposed algorithm.2. A robust watermarking scheme is proposed against geometric attacks, which is combined the feature points with the gradient directions. Because the gradient direction angle between two feature points still remains unchanged under geometric attacks, watermark can be embedded by modifying it. And the embedding can be theoretically completed by modifying the neighborhood DCT coefficients of feature points. The algorithm does not normalize a part of the image or the whole, which avoids the normalization out-synchronization caused by the shearing and cropping attacks. Experimental results show that the algorithm can resist many image processing and geometric attacks, especially for shearing and cropping attacks.3. A self-embedding fragile watermarking algorithm based on run-length coding is proposed to eliminate the block independence due to the block image authentication algorithm, and further reduce the impact of the self-embedded watermark on the carrier quality. In the algorithm, the original image is firstly divided into2x2blocks, and create a thumbnail by the each sub-block gray average values. Then, encoding is made by the relativity between pixels of the thumbnail. And the code is embedded as watermark into LSB of image. While detecting, watermark is matched with image block according to the string match. If blocks are unsuccessfully matched, they are re-matched by the group to complete the authentication and recovery. The theoretic analysis and the simulation results show that it can not only resist the collage attack and locate tampered blocks accurately, but also has better non-visibility.4. A self-embedding fragile watermarking algorithm is proposed based on the joint decision of the offset sub-block to locate and recover the randomly tampered areas. In this scheme, image is firstly blocked into with the size of2x2. And then the content features are encoded into5bits as watermarks, which are embedded into the mapping block. While detecting, content features and watermark are matched with those of its mapping blocks, which determine the tampered blocks. During restoring watermark, the watermark credibility is determined by whether its offset sub-block is certified. The theoretical analysis and experiments make clear that, the proposed scheme not only has lower missing rate but also put up better performance for randomly tampering.5. A JPEG block authentication algorithm is proposed by combining matrix encoding with the multiple watermarking. Matrix encoding can shorten the length l of watermark into log2l, and accurately locate one bit changed information. And then multi-watermarks are generated to locate the tampering. The detailed analysis is given about the missing rate, the false detection rate and watermark visibility in this paper. In addition, how to choose parameters is also shown in the different applications. Theoretical analysis and experimental results show that JPEG block-level tamper localization can be done under the lower tamper rate. And more than that the missing rate and the false detection rate are at the10-2level, even if only lbits watermark are embedded and the rate of tampering is less than5%.6. A multifunctional watermarking algorithm is proposed for image integrity protections. The original image is divided into non-overlapped blocks with size of3×3. The robust watermark is embedded into the gradient directions of the center pixels. And the fragile watermarking is embedded into the DCT coefficients. The joint decision of the offset sub-block is used to locate and recover the randomly tampered areas. Meanwhile, the types of attacks are discriminated according to the normalized correlation of the watermark and the extracted watermark. The simulation results show that the embedding and extraction of the robust watermark and fragile watermark are independent to each other. Moreover, this algorithm not only can locate and recover the tampered areas, but also can discriminate the types of attacks.Finally, a conclusion is given and the next future researches are also discussed in the thesis.