Identifying Critical Infrastructure Clusters via Spectral Analysis

In this paper we discuss how to identify groups composed of strictly dependent infrastructures or subsystems. To this end we suggest the use of spectral clustering methodologies, which allow to partition a set of elements in groups with strong intra-group connections and loose inter-group coupling. Moreover, the methodology allows to calculate in an automatic way a suitable number of subsets in which the network can be decomposed. The method has been applied to the Italian situation to identify, on the base of the Inoperability Input-Output model, which are the most relevant set of infrastructures. The same approach has been applied also to partition in a suitable way a network, as illustrated with respect to the IEEE 118 Bus Test Case electric grid.