Reversible fragile watermarking for locating tampered blocks in JPEG images

This paper proposes a novel fragile watermarking scheme for JPEG image authentication. The watermark is generated by folding the hash results of quantized coefficients, and each block is used to carry two watermark bits using a reversible data-hiding method. Because modification to the cover is small, the visual quality of watermarked image is satisfactory. On the receiver side, one may attempt to extract the watermark and recover the original content. By measuring mismatch between the watermark data extracted from the received image and derived from the recovered content, the blocks containing fake content can be located accurately, while the original information in the other blocks is retrieved without any error as long as the tampered area is not extensive.