Distortion Optimization of Model-Based Secure Embedding Schemes for Data-Hiding

This paper is the continuation of works about analysis of secure watermarking schemes in the case of WOA (Watermarked Only Attack) framework. In previous works, two new BPSK spread-spectrum watermarking modulations, Natural Watermarking (NW) and Circular Watermarking (CW), have been proposed and have been shown to be more secure than classical modulations. Because security is guaranted using specific distributions of watermarked contents, we propose to use optimal model-based embedding to ensure security while minimizing the overall distortion. Additionally, we propose a new secure watermarking scheme based on distribution of vector norms in the Gaussian case. We illustrate model-based embedding performance in the case of Gaussian signals and show that this approach not only allows to achieve excellent level of security in the WOA framework, but also allows to minimize distortion. Finally, a comparison of the robustness of the proposed embedding schemes is performed.