The Stackelberg equilibrium for one-sided zero-sum partially observable stochastic games

Formulating cyber-security problems with attackers and defenders as a partially observable stochastic game has become trend recently. Among them, the one-sided two-player zero-sum (OTZ-POSG) emerged popular model because it allows players to compete for multiple stages based on partial knowledge of system. All existing work OTZ-POSG focused simultaneous move scenario assumed that one player’s actions are private in execution process. However, this assumption may questionable since action be detected by opponent through deploying detection strategies. Hence, paper, we propose turn-based public investigate existence properties Stackelberg equilibrium game. We first prove one-stage case show can converted into linear-fractional programming problem therefore solved linear programming. For stages, main challenge is information leakage issue run-time reveals certain achieve more rewards future. To deal issue, adopt concept ε-Stackelberg achieved finite-horizon OTZ-POSGs. space partition approach solve iteratively value function leader piece-wise follower constant stages. Finally, examples given illustrate functions constant.