Input-output Inoperability Risk Model and Beyond: A Holistic Approach 4 views

Assessing and managing risks inherent in a set of large-scale, complex engineered systems (such as critical infrastructures) due to the intra- and interdependency are important from both theoretical and practical standpoints. In this research, we lay out a framework based on the notion of Input-Output that can help describe and manage such risks as well as their propagation and proliferation. Our fundamental assumption is that the total risk, expressed in terms of equilibrium in operability!, is the joint effect of an initial perturbation and the interdependency inherent in the system. This joint effect can result in catastrophic consequences and very long recovery time. The overall goal is accomplished through completion of the following: 1. Develop a general methodology for identifying sources of risks confronted by large-scale, complex engineered systems. This process builds on the philosophy and methodology of the Hierarchical Holographic Modeling. It is proposed that the risk identification process is marked by three steps and the final product is characterized by a combinatorial mapping of the hazard HHM onto the system HHM. 2. Develop a perturbation-based inoperability risk model for interdependent complex systems based on Leontief s concept of input and output and earlier work by Haimes and Jiang. The notion of inoperability is further developed as a general risk metric to enable a unified conceptual approach in the modeling framework. The notion of derivative inoperability is introduced to capture the consequences of an initial structural inoperability resulting from an attack. This model serves as a linear approximation of a more general risk model for small initial perturbations. Furthermore, the model is capable of providing an assessment of both the lower and upper bounds of the equilibrium inoperability. Both the capabilities and the limitations of the proposed model are investigated. 3. Study the dynamics of risk due to an initial perturbation. Initial structural inoperability can propagate to other systems in the form of derivative inoperability by reducing support (result of serial dependency) or by overloading (result of parallel dependency). The concepts of influence set and reliance set are introduced to describe system connectivity. Based upon these concepts, inoperability influence diagram is used to show the propagation of inoperability over time triggered by an initial disturbance, and inoperability proliferation factor is used to capture the increase in inoperability in a truly bidirectional system. Some general theoretical results are obtained. Abstract models in both discrete and continuous cases are investigated. This study sheds light on the temporal pattern of cascading effects resulting from an attack. 4. Develop a risk management methodology that aims at minimizing the derivative inoperability of a set of serially interdependent systems according to prioritized objectives. Using optimization techniques, it is demonstrated that the derivative inoperability can be significantly reduced by deliberately distributing the initial inoperability to other systems so that the total loss (or inoperability) is minimized. The optimal distribution strategy is found by linear programming. We applied this methodology to a Leontief-based economic system, through a case featuring 12 economic sectors where derivative inoperability is manifested as derivative economic loss. It is shown that the total economic loss is significantly reduced by applying the risk management strategy. One important policy implication of this result is that the public can be better served by the redistribution of risk.

PHD (Doctor of Philosophy)

Risk management; Cyberinfrastructure; Internet--Security measures

English

All rights reserved by the author (no additional license for public reuse)

2003-08-01