Research Of Image Fragile Watermarking Algorithms

As one of the major media in the information age, digital image is playing an extremely important role in the transmission of message. Due to unprecedented openness of current network and increasing of computing capability, any adversary can easily tamper images without arousing suspicion. So, it is hard to distinguish if the image comes from the certain source, and any misuse of uncertificated image will generate unexpected consequences. Image fragile watermarking, as one of the main branch of digital watermarking, can solve this problem. It can not only identify the image content integrity and authenticity, but also can localize or recover the tampered region.In image fragile watermarking technique, the algorithm's security, the accuracy of localization and the capability of recovering tampered region are the main areas of research. In this paper, taking BMP image as cover, a secure self-embedding algorithm is proposed. Three secret keys are taking as the inputs of secure Hash function, and the embedding location of recovering watermark of each block in the image is selected pseudo-randomly using the output of secure Hash function. Theoretically, the inputs of secure Hash function can be any combining of numbers or letters, so the key space of the proposed algorithm is infinite. Furthermore, taking the coefficients of integer DCT transform as watermark, the recovering capability is better than that of traditional DCT transform.In addition, the security of an image authentication algorithm based on fragile watermarking is analyzed in this paper. To overcome these drawbacks, a successful attack is constructed:given a watermarked image, the selected regions are tampered. Secondly, the watermark bits for tamper localization of size 4×4 and 2×2 block in the tampered regions are forged, and then the forged watermark is embedded according to the embedding algorithm. This attack will lead to that the localization of tampered image blocks is inefficient. So, in this paper, an improved algorithm is proposed. With the help of Hash function, the offset blocks are selected randomly. And the watermarks of blocks with the size of 4x4 and 2x2 are embedded into corresponding offset blocks. Without the knowledge of secret keys, any adversary cannot tamper the given image and forge the watermark synchronously. Furthermore, the content-based recovery watermark is embedded into the cover image without arousing any noticed distortion. So, if the detected image is forged, the recovery watermark can be extracted to recover the tampered regions. These improvements make the algorithm more suitable for practical application.