How to combat block replacement attacks?

Block replacement attacks consist in exploiting the redundancy of the host signal to replace each signal block with another one or a combination of other ones. Such an attacking strategy has been recognized to be a major threat against watermarking systems e.g. additive spread-spectrum and quantization index modulation algorithms. In this paper, a novel embedding strategy will be introduced to circumvent this attack. The basic idea is to make the watermark inherit the self-similarities from the host signal. This can be achieved by imposing a linear structure on the watermark in a feature space e.g. the Gabor space. The relationship with existing multiplicative watermarking schemes will also be exhibited. Finally, experimental results will be presented and directions for future work will be discussed.