Game-Theoretic Approach to Attack Planning and Controller Placement in Software Defined Networks

The paper addresses a game-theoretic approach to controller placement and targeted attack planning problems. The two models are presented. The first model, the attacker's point of view, allows for efficient attack planning using mixed strategies to maximize damage in the network (minimize expected availability measure). The second model, the operator's point of view, allows for efficient controller placement using mixed strategies to maximize the expected availability measure and to defend against mixed-strategy attacks. The assumption is that attacks and placements occur in multiple rounds in the form of a game. Within Software Defined Network (SDN), such a game can technically be conducted using network programmability since the controllers may be relocated. The two mixed-strategy models were implemented and tested using linear programming solvers and compared to the min-max and max-min models (single best strategy), respectively.