Identifying vulnerabilities and extreme risks in critical infrastructure networks

Abstract

Today’s society is becoming highly dependent on the services provided by various critical infrastructure networks. The increasing complexities and interdependencies among infrastructure networks have exacerbated their susceptibility to various disruption or failure events. It is therefore crucial to understand how the failure of the components in a critical infrastructure network affects the performance and integrity of the whole network. Different types of component failures may result in different levels of failure consequence and it is interesting to investigate which type of failure results in the largest failure consequence in an infrastructure network. A review on critical infrastructure protection shows that there is a need to incorporate geographic proximities in the failure cascading process in infrastructure networks. Hence this research initially investigates the failure consequence in a critical infrastructure network resulting from different types of component removals using a proximity-based cascading model for failure propagation. The feasibility of the proposed study is tested on a real-world transportation network. A review on critical infrastructure simulation and analysis also suggests that all the analyses and subsequent policy decisions on critical infrastructure networks have been made based on the assumption that the infrastructure interdependencies model has been constructed to a fair degree of completeness. Although such analyses that aim at the identification of weaknesses and vulnerabilities in infrastructure networks may go some way in preventing or alleviating disastrous outcomes, the failure to consider unforeseen interdependencies among critical infrastructures can result in extreme disruptions not being anticipated. The review also indicates that although approaches for vulnerability/consequence analysis have been widely studied, a complete risk assessment of infrastructure network disruptions incorporating the probabilities as well as consequences of disruption events has not been given serious attention. Therefore this research also proposes using an optimization algorithm to iteratively search for the possible unforeseen interdependencies as well as the failure points that can result in extreme risk in critical infrastructures, thereby anticipating extreme risk events. In order to illustrate the feasibility of the proposed approach, an agent based model of an infrastructure network along with its known interdependencies has been presented, with a genetic algorithm applied to search for potential unforeseen interdependencies as well as failure modes/points that can result in extreme disruptions. The results from this study show the feasibility of anticipating extreme risk events in infrastructure networks, thereby providing valuable insights for proactive risk management of critical infrastructures.