Model-Driven Estimation of Distributed Vulnerability in Complex Railway Networks

The high level of interconnection and interdependency between critical infrastructures demands for novel methodologies and techniques to improve their resilience by evaluating the vulnerability of these systems in presence of attacks. Railway networks, as all other critical infrastructures (electrical, communication, etc.), are highly interconnected. One of the most valuable tasks in designing protection systems is detecting the critical nodes. This paper introduces a model-driven approach in order to accomplish such resilience analysis. The network is modeled by using the CIP_VAM UML profile that provides a way to model complex cyber-physical systems in a unifying manner with respect to services, components, attacks and protections: a special focus is on the modeling of the inter-dependencies between infrastructures. Then the vulnerability analysis is performed by generating a Bayesian Network model capable to take into account probabilistic relationships between nodes that translate services, components, attacks and protections. Such approach allows for sensitivity analyses in order to detect vulnerability flaws for an effective and efficient protection improvement.