CNN-based steganalysis and parametric adversarial embedding:A game-theoretic framework

Highlights：

• Introduce two game theoretic frameworks of the Adversary-aware Stego Embedding/Detection game, derive the equilibrium points and the corresponding payoff for the two versions of the game.

• The game in the Bayesian decision setup, non zero-sum game can be traced back to zero-sum game, for which the Nash equilibrium of a zero-sum game is significantly easier to be derived.

• The optimum strategies of addressing the interplay between the steganographer and the steganalyst in a game-theoretic fashion come out when compare to use other strategies.

• Steganalyst can still improve the performance of steganalyzer by training it with a mixture of images distributed according to steganographer’s distribution at the mixed Nash equilibrium strategy.