Abstract-
I. INTRODUCTIONUnderstanding, modeling, and assessing the normal functioning and the possible faulty conditions of critical infrastructures (CIs) is essential to safely design these complex systems, and to effectively operate the services they provide. Distributed infrastructures have been modeled as graphs to unravel their structure and dynamics, and to evaluate to what extent the structure may impact the dynamics [1]. However, it is evident that infrastructures do not 3 exist in isolation of one another, and the relations among them must be identified to perform realistic and applicable analyses [2]. The focus of the research on critical infrastructures must then be shifted from single, isolated systems to multiple, interconnected, mutually dependent systems. The aim becomes that of assessing the influences and limitations which interacting infrastructures impose on the individual system operating conditions, for avoiding fault propagation by designing redundancies and alternative modes of operations, and by detecting and recognizing threats [3].In modern society, the linking among service infrastructures is required for optimal and economical operation.Yet, these interconnections introduce weaknesses in the systems due to the fact that failures may cascade from one system to other interdependent systems until, possibly, affecting the overall functioning, if proper protection of interdependencies is not considered [4]. The role of dependencies among infrastructures (so called interdependencies), and the intrinsic difficulties arising in their modeling, have been highlighted in empirical studies. For example, the database in [5] has been built from public reports of disruptions of CIs from open sources like newspapers and internet news. Events have been classified as "cascade initiating" (i.e., an event that causes an event in another CI), "cascade resulting" (i.e., an event that results from an event in another CI), and "independent" (i.e., an event that is neither a cascade initiating nor a cascade resulting event). The information in the database shows in particular that: i) "cascade resulting" events are more frequent than generally believed, and that "cascade initiators" are about half as frequent; ii) the dependencies are more focused and directional than often thought; and iii) energy and telecommunication are the main "cascading initiating" sectors.Interdependencies among telecommunication systems, transportation systems, and power distribution grids played a negative role in the small telecommunication blackout which took place in the suburbs of Rome in January 2004[6]. The chain of events originated in a major telecommunication service node when a metallic pipe carrying cooling water for the air conditioning broke. The resulting flood led to a communication blackout in the area, which in turn hit the country's biggest printed news agency transmissions; stopped the check-in, ticketing and luggage acceptance at the International Fiumicino Airport; disturbed post offices and banks operations; a...
