Optimal protection of general source-sink networks via evolutionary techniques

This paper presents a network optimization model that maximizes the survivability of a network for a specific source-sink flow when resources are allocated to protect network links under an evenly distributed attack strategy. The proposed model takes into account the stochastic behavior of link interdiction by assuming that the vulnerability of an attacked/defended link is determined via a ratio form of the attacker-defender contest success function. The research done as part of this manuscript extends previous network interdiction approaches through an optimization model that relates the defenders'/attackers' per-link resources with link vulnerability values allowing for the defender to consider competing courses of action. Once the network vulnerability model is proposed, the manuscript describes an evolutionary algorithm that can provide optimal defense strategies in networks that follow a source-sink behavior. The results of the extensive experimentation done for different sized networks illustrate that when a defender is faced with an evenly distributed attack his/her defense strategy should take into account the configuration of the network, the source-sink flow and the importance of the different network elements.