oa The Infrastructure Of Critical Infrastructure: Vulnerability And Reliability Of Complex Networks

* Background & Objectives All critical infrastructure can be modeled as networks, or systems of nodes and connections, and many systems such as the electric grid, water supply, or telecommunications exist explicitly as networks. Infrastructures are interdependent, for instance, telecommunications depend on electric power, and control of the electric grid depends increasingly upon telecommunications, creating the possibility for a negative feedback loop following a disturbance. The performance of these systems under disturbance are related to their inherent network characteristics, and network architecture plays a fundamental role in reliability. What characteristics of networks affect their robustness? Could, for instance, the vulnerability of the electric grid to cascading failure be reduced? * Methods We create a failure model of the network where each node and connection is initially in the operative state. At the first discrete time step a network element is changed to the failed state. At subsequent time steps a rule is applied which determines the state of random network elements based upon the state of their neighbors. Depending upon the rule and the distribution of the degree of connectedness of the network element, failures may be contained to a few nodes or connections, or may cascade until the entire network fails. * Results Quantitative measures from the model are the probability of network failure based upon the loss of a network element, and the expected size distribution of failure cascades. Additionally, there is a critical threshold below which infrastructure networks fail catastrophically. The electrical grid is especially vulnerable as it operates close to the stability limit, and there is a low critical threshold after which the network displays a sharp transition to a fragmented state. Failures in the electrical grid result not only in the loss of capacity in the network element itself, but load shifting to adjacent network elements, which contributes to further instability. While most failures are small, failure distributions are heavy tailed indicating occasional catastrophic failure. Many critical infrastructure networks are robust to random failure, but the existence of highly connected hubs give them a high clustering coefficient which makes the network vulnerable to targeted attacks. * Conclusions It is possible to design network architectures which are robust to two different conditions: random failure and targeted attack. It is also possible to alter architecture to increase the critical threshold at which failed network elements cause failure of the network as a whole. Surprisingly, adding more connections or capacity sometimes reduces robustness by creating more routes for failure to propagate. Qatar is in an ideal position to analyze and improve critical infrastructure from a systemic perspective. Modeling and simulation as detailed above are readily applicable to analyzing real infrastructure networks.