Identifying Optimal Attacks for Infrastructure Networks with failure Propagation

Abstract

Most critical infrastructures can be represented as directed graphs in which the nodes denote the operational entities and the arcs denote the dependencies of one entity upon others. A rational adversary that plans to harm the system would aim to use the minimum amount of effort to cause the maximum amount of damage. Therefore, attacking the most critical nodes of the graph would be the most rational strategy of an attacker. In contrast, the system's operator might be able to minimize the damage if they are able to defend these critical nodes. Hence, it is strategically important for both the attacker and the defender to identify these critical nodes, but this corresponds to a complex optimization problem. This paper presents a problem formulation and then a solution for computing the optimal cost attack that takes failure propagation into consideration. Given a graph representing one or multiple infrastructures and the attack costs of each node, the proposed method computes the optimal attack that a rational attacker is likely to make. Our problem model has the goal of maximizing the damage for a given attack budget. Our technique relies on integer-linear programming to obtain the solution, given our objective function and sets of constraints for each of the specified goals. As shown by the experimental results, using a wide range of graphs, the proposed solution is very fast in terms of the runtime complexity.