Multi-objective evolutionary optimisation of the protection of complex networks exposed to terrorist hazard

Abstract

This paper extends previous works developed for evaluating the security characteristics of a network system exposed to a terrorist attack (e.g. the injection of poison in the water distribution system of a populated area), by defining a robust protective scheme based on a Multi-objective Optimisation (MO) approach. The physical system can be modelled as a network (graph) of nodes interconnected by links. Two criteria are used to measure the vulnerability of the network: the Time To Reach All network Destination nodes (TTRAD) and the Average Number of Persons Affected (ANPA) which measures the damage that the hazard causes. The optimal system protection is tackled using a Multi-Objective Evolutionary Algorithm (MOEA) which is part of a family of evolutionary algorithms tailored to cope with MO problems. The goodness of each alternative protection scheme is quantified by a combination of Cellular Automata and Monte Carlo simulation. Numerical examples illustrate the approach.