Compressed-domain fragile watermarking scheme for distinguishing tampers on image content or watermark

Nowadays the compressed format files such as JPEG are used widely, but the existing compressed-domain fragile watermarking algorithms can not recognize whether the modification made to the watermarked image is on the image contents or the embedded watermark or both tampered by attackers. In this paper, a compressed-domain fragile watermarking scheme with discrimination of tampers on image content or watermark is proposed. The quantized DCT (discrete cosine transform) coefficients of 7 most significant bits (MSBs) after decoding are mapped to the initial values of the chaotic system, and then generate content-based fragile watermark by chaotic motion. After watermark permutation by the chaotic sequence, the operations of embedding and extracting the improved security watermark are all in the compressed domain. During authentication, the proposed method is able to detect the tampered location and discriminate tampers on image content or watermark from the distribution density of nonzero elements in the tamper detection matrix. The quantized DCT coefficients after decoding is used directly for watermark embedding, which avoids a large of computation on full decoding and re-encoding process. Therefore, the proposed algorithm has a great efficiency when it is implemented. Furthermore, thanks to the high sensitivity on initial values and the good randomness of the chaotic system, the very accurate localization is not only realized for the malicious tampers to JPEG images, but also indicate whether the modification made to the watermarked image is on the image contents or the embedded watermark or both. Simulation results show the effectiveness, sensitivity and ability of tamper localization of this technique.