Single Plant Protection: Playing the Chemical Plant Protection Game Involving Attackers with Bounded Rationality

In this chapter, we model attackers with bounded rationality in the Chemical Plant Protection game. Three different behaviour models of attackers are investigated, namely, the epsilon-optimal attacker, the monotonic-optimal attacker, and the MiniMax attacker. All these attacker models are integrated to the Stackelberg CPP game, which means that the defender moves first, and the attackers follow. Furthermore, the monotonic-optimal attacker is investigated in the Interval CPP game with only one type of attacker, and a game solution named Monotoic MaxiMin Solution for the Interval CPP game (MoSICP) is defined [1]. The MoSICP solution incorporates both bounded rational attackers and distribution-free uncertainties into the CPP game. The epsilon-optimal attacker model, being related to the defender’s distribution-free uncertainties, and the MiniMax attacker model, being the most conservative model, are therefore investigated in the Bayesian Stackelberg CPP game framework, instead of in the Interval CPP game framework. The defender is still assumed to behave rationally to maximize her payoff.