Reliability evaluation by expected achievable capacity in stochastic network using game theory

To obtain a network with high fault tolerance, all possible characteristics of a failure event must be captured in the analysis. Also, an efficient method to identify and then upgrade vulnerable network components is required. A network game model between a router and intelligent attacker has been widely explored to overcome this challenge. In this paper, based on game theory framework, we have proposed a new vulnerability identification method to measure network reliability when the network is attacked by an intelligent adversary, who destroys network links to minimize capacity achieved between two network terminals. A new performance indicator-expected achievable capacity (EAC) has been proposed to help quantifying link vulnerability. To obtain EAC, a maximin problem is formulated and the method of successive averages is chosen to solve for the game solution. Numerical results show that the effect of worst-case failure on EAC can be thoroughly analyzed by the proposed framework.