Research On Image Content Authentication And Recovery Based On Self-embedding Technology

The rapid developed mobile internet provides a more efficient channel for the transmission and acquisition of digital multimedia.Under such an open environment,a series of attacks such as malicious tamper,delete and copy will happen during the transmission and application processes of digital multimedia,which inevitably destroy its integrity and security.While digital watermarking technique is an effective method to solve this problem and it can effectively realize the copyright protection and content authentication of digital media.Especially,the self-embedding watermarking technology,which can not only authenticate the integrity of multimedia,but also locate the tampered region and recover the tampered area,has become a hot topic in the field of information security.In this paper,the digital watermarking algorithm based on the self-embedding technology has been studied with the aim to realize the content authentication and recovery of digital image effectively.On the basis of summarizing the advantages and disadvantages of the existing digital image watermarking algorithms,three new watermarking algorithms based on the self-embedding technique were proposed:Algorithm one proposed a robust watermarking algorithm based on the self-embedding technique with totally blind detection function.The algorithm combined two-dimensional discrete cosine transform(DCT)and singular value decomposition(SVD),from which a unique feature watermark was generated by the two transform domain coefficients,and then the watermark was embedded into the image.The algorithm realized totally blind detection by calculating the correlation coefficient between the generated feature watermark and the extracted authentication watermark.The experimental results show that the proposed watermarking algorithm has strong robustness to resist various common image processing attacks and can keep good invisibility of the embedded watermark.Algorithm two proposed a totally blind and multi-functional digital watermarking algorithm based on the self-embedding technology.The algorithm adopted the idea of region segmentation for sub-block,and the generation and embedding of watermark were respectively carried out in the different regions of the same sub-block.A feature watermark was generated by judging the parity of the first digit from the average of maximum singular values,and then the watermark was embedded by adjusting two DCT coefficients of adjacent blocks in the same position.Both the image copyright protection and content authentication functions were achieved by comparing the generated feature watermark and extracted authentication watermark in the detection.The experimental results show that the proposed algorithm can not only realize the copyright protection and content authentication,but also locate the tampered area and distinguish the type of tampering.Algorithm three proposed a self-embedding digital watermarking algorithm based on the singular value decomposition for image tamper detection and recovery.The algorithm created an authentication watermark on the singular value decomposition of the image blocks and generated a recovered watermark on the mean value of the image blocks.The two bits authentication watermark was embedded into its own block by cross-embedding method,and meanwhile part of the authentication and recovery watermarks were embedded into the mapping block.The tamper detection of digital image was realized by comparing the extracted authentication watermark and the recovered watermark.The experimental results show that the proposed algorithm can not only locate and recover the tampered area accurately,but also resist the forgery attacks such as the collage attack,content-only tampering and constant-average attack.