EWA: An exemplar-based watermarking attack

Abstract

In this paper, we used image inpainting as a means for reconstruction of small damaged portions of a watermarked image so that the hidden watermark is not detectable afterwards. Our new watermarking attack algorithm extends the exemplar-based inpainting method, and then we apply it to intended noisy watermarked images to alter the watermark detection process. Our approach is completely free from any pre-assumption on the watermarking algorithm or any other parameters that is used during the watermark embedding procedure. The average PSNR/SSIM of the watermarked image after applying the proposed attack is more than 35/0.98 and the average NC for extracted watermark is lowers than 0.5, so theoretically, the watermark is not detectable.