Attack on Cocktail Watermarking Based on High False Positive Probability

In this paper, a simple and effective attack on a non-blind robust digital watermark method, cocktail watermarking, was proposed. After reviewing the principle of the cocktail watermarking, the paper discovered that the cocktail watermarking scheme can drive a very high false positive probability although with high robustness simultaneously. Furthermore, a counterfeiting attack to generate a watermarked image with an original image was proposed, which can pass through the watermark extraction and watermark authentication successfully in the case of none watermark embedded. This attack scheme demonstrates that the watermarking algorithm is flawed which can be used by the malicious content owner to forge and distribute a watermarked image to the unknown users. At last, simulations of the counterfeiting attack on cocktail watermarking are devised and the results show that this attack is successful with a reasonable image quality.