A realistic attack on SVD based watermarking scheme

Abstract

In this paper we propose a man in the middle attack on a class of SVD based watermarking methods. We have shown that our attack is successful in creating false positive detection rate with probability one. We also show the previously proposed a man in the middle attack is not realistic. Consequently, our proposed attack is the first realistic attack on this class of watermarking methods. The simulation results confirm the practicality of our proposed attack in creating false positive detection in watermark extraction. Also, the simulation results show our proposed attack is able to extract any desired watermark from the image watermarked by the SVD method.