Interdiction-Based Approaches to Identify Damage in Disrupted Critical Infrastructures with Dependencies

Abstract

In this paper, we consider a problem in which the decision maker only has access to outage reports from service receivers in critical infrastructures with dependencies among them, and our objective is to determine inspection plans to identify the damaged components that lead to the reported outages. This situation could occur, for example, if a cyberattack on the human-machine interface of the infrastructures occurs at the same time a natural hazard damages physical components. The demand nodes within these infrastructure networks may have an outage either from damage that occurred within their infrastructure or due to cascading failures across infrastructures. We present network interdiction-based approaches to identify a set of components to inspect that replicates the service outages reported across the infrastructures. These approaches can be applied in an iterative nature where a particular approach first identifies a set of components across infrastructures to inspect, and then, based on the feedback received from the inspections, determines an updated set of components to inspect. Performance criteria of the proposed approaches are the percentage of successful inspections, which represents the percentage of inspection resources used to successfully identify damaged components. As a benchmark, we applied a breadth-first search (BFS) approach to the problem, which mimics a top-down inspection policy for identifying damaged components within an infrastructure. We tested our approaches on realistic infrastructure systems and demonstrated that they far outperform the BFS approach.