Empirical Characterization of Infrastructure Networks [INCCA]
Critical infrastructure protection is a recognized problem of national importance. Infrastructure networks such as electric power, natural gas, communications, and transportation systems have an inherent graph-theoretic structure. Quantitatively characterizing the essential properties of infrastructure networks for various domains lays a valuable foundation for studying the universal features (especially criticality, robustness, etc.) and specific characteristics of such networks. We construct an extensive reference data set of infrastructure network graphs: 44 graphs of 13 types with nearly one million vertices and over one million edges. After regularizing these graphs, we compute more than fifty metrics related to connectivity, distance scale, cyclicity, cliquishness, and redundancy. We contrast these metrics for different types of infrastructures, study their interrelationship, and use them to cluster and classify systems. We consider both intact networks and networks that have been degraded by the removal of some vertices or edges either at random or systematically—this provides insight as to the robustness of the network if it were subject to a natural disaster or an attack.
Selected Publications
B. W. Bush, C. R. Files, and D. R. Thompson, “Empirical Characterization of Infrastructure Networks,” Los Alamos National Laboratory, Report LA-UR-01-5784.
Critical infrastructure protection is a recognized problem of national importance. Infrastructure networks such as electric power, natural gas, communications, and transportation systems have an inherent graph-theoretic structure. Quantitatively characterizing the essential properties of infrastructure networks for various domains lays a valuable foundation for studying the universal features (especially criticality, robustness, etc.) and specific characteristics of such networks. We construct an extensive reference data set of infrastructure network graphs: 44 graphs of 13 types with nearly one million vertices and over one million edges. After regularizing these graphs, we compute more than fifty metrics related to connectivity, distance scale, cyclicity, cliquishness, and redundancy. We contrast these metrics for different types of infrastructures, study their interrelationship, and use them to cluster and classify systems. We consider both intact networks and networks that have been degraded by the removal of some vertices or edges either at random or systematically–this provides insight as to the robustness of the network if it were subject to a natural disaster or an attack.