Attacks on a robust watermarking scheme based on self-reference image

In this paper, we show that the recent robust watermarking scheme by Liu et al. is insecure and does not provide robustness. In particular, the embedding space where the watermark is located can be recovered, and therefore removed or overwritten. This means the scheme fails when used to provide robustness applications like proof of ownership and copyright protection. The conventional approach to judging the security of a scheme is that when it has been published for some time and yet no successful attacks exist, the scheme is taken to be a secure design; thus will be possibly implemented by researchers or the industry. It is therefore important to highlight these attacks otherwise the scheme may be implemented by people unaware that such attacks exist.