Watermark Attack Based on Blind Estimation without Priors

Abstract

With the rapid growth of multimedia application, there has been significantly bared a crucial need for effective copyright protection mechanisms. Therefore, digital image, video, audio and text watermarking has recently become a very active research area of signal and information processing, as a solution to this problem. Many important issues have been pointed out, one of them being the robustness to non-intentional and intentional attacks. In this paper, a new estimation based watermark attack method is presented. The proposed attack regards the watermark system as a black box by assuming watermark signal as different kinds of noise, instead of using the prior knowledge of the watermark signal and embedding methods. To estimate the original image data, blind estimation and a MIMO linear channel are introduced. Experimental results demonstrate the proposed watermark attack method show the high efficiency against known academic and commercial methods. The watermark is completely destroyed in all tested images without altering the image quality.