“Identifying Vulnerable Systems and Understanding the Effects of Network Properties on the Robustness of Interdependent Systems”

Thu Mar 21, 2019 2:00 PM

Location: LTS Auditorium, 8080 Greenmead Drive

Richard J. La
Professor, Department of Electrical and Computer Engineering and Institute for Systems Research

Modern societies are becoming increasingly reliant on complex systems to deliver vital services, e.g., electrical grids, transportation systems. As the security and operations of subsystems in these systems are interdependent, they are dubbed interdependent systems. Due to the growing complexity of these interdependent systems and intricate interactions and dependence among comprising subsystems, analyzing them is a major challenge that researchers and engineers face today and little is known about their robustness or vulnerability to failures and attacks. In this talk, we will discuss two key issues that arise in understanding and strengthening the robustness of such critical systems against failures and attacks.

First, we study the problem of identifying subsystems that are more prone to trigger widespread failures throughout the system, e.g., a large-scale blackout in a power grid, in the event of their failures and hence should be reinforced. We call such subsystems “vulnerable” subsystems. We propose two new computationally efficient approaches to finding such vulnerable subsystems and show that, under some assumptions, there is an interesting relationship between the likelihood of a subsystem setting off a potentially catastrophic widespread failures and its centrality in the system. With the help of numerical studies, we demonstrate that the proposed approaches are effective at identifying most vulnerable systems.

Second, we examine the influence of graph properties, such as clustering and degree correlations among others, on the robustness of interdependent systems. We show that increasing clustering in the system hampers cascading failures among subsystems, making it less likely to suffer widespread failures throughout the system. In addition, we demonstrate that, when strategic decision makers choose the security investments for subsystems in order to protect them against attacks and failures, the effects of degree correlations depend on the cost-effectiveness of available security measures: if the security measures are cost-effective, higher degree correlations in the system tend to improve the security and robustness of the overall system, whereas the opposite is true when the security measures are not cost-effective.

Speaker Bio:
Richard J. La is a professor in the Department of Electrial and Computer Engineering and the Institute for Systems Research at the University of Maryland.

He is currently an associate editor for IEEE/ACM Transactions on Networking. La served as an associate editor for IEEE Transactions on Information Theory and IEEE Transactions on Mobile Computing as well as an editor for IEEE Communications Surveys and Tutorials.

He received his doctorate in electrical engineering from the University of California, Berkeley in 2000.