Information-theoretic resolution of perceptual WSS watermarking of non i.i.d. Gaussian signals

TL;DR

This paper extends the theoretical framework of watermarking by analyzing non-i.i.d. Gaussian signals with weighted distortion, providing closed-form solutions for optimal embedding and attacks in spread spectrum schemes.

Contribution

It introduces a generalized information-theoretic approach to perceptual watermarking for non-i.i.d. signals, including a game-theoretic analysis of optimal attacks and practical embedding formulas.

Findings

Closed-form expressions for watermark embedding parameters

Generalized attack strategies derived from game theory

Validation of the approach for non-i.i.d. Gaussian signals

Abstract

The theoretical foundations of data hiding have been revealed by formulating the problem as message communication over a noisy channel. We revisit the problem in light of a more general characterization of the watermark channel and of weighted distortion measures. Considering spread spectrum based information hiding, we release the usual assumption of an i.i.d. cover signal. The game-theoretic resolution of the problem reveals a generalized characterization of optimum attacks. The paper then derives closed-form expressions for the different parameters exhibiting a practical embedding and extraction technique.