Title:Information Leakage Games

Abstract:We formalize the interplay between defender and adversary in a game-theoretic framework adapted to the specific issues of quantitative information flow. Assuming that both defender and adversary may be active and influence the system during the attack, we define a general framework of information leakage games in which the payoff function of the game is information leakage. We provide methods for finding the solution and the equilibria of various kinds of leakage games, varying according to whether players act simultaneously or sequentially, and to whether the defender can conceal from the adversary the channel used. We demonstrate that in some cases it is in the best interest of not only the defender, but also of the adversary, to employ randomization in the choice of actions. We illustrate the power of our framework with a detailed case study of the Crowds protocol running on a MANET (Mobile Ad-hoc NETwork).