2010 Miura-Ko PhD Dissertation Modeling and Mitigation of IT Risks Stanford
The dissertation develops quantitative, decision‑theoretic models for managing information‑technology (IT) security risk. It begins by surveying existing risk‑measurement frameworks (e.g., ALE, attack trees) and then introduces three original modeling contributions. The first, SecureRank, is a risk‑based vulnerability‑prioritization scheme that ranks patches by their expected impact on overall system risk, accounting for network topology and inter‑node dependencies. The second contribution formulates interdependent security investment games, where multiple organizations choose security expenditures that generate both positive and negative externalities; the analysis derives equilibrium strategies and explores the effects of network structure, diagonal dominance, and subsidy mechanisms. The third contribution presents a dynamic risk‑network model that treats inherent risk at each node as a stochastic process evolving in discrete (or continuous) time, with defense modes influencing transition probabilities. Bellman equations and dynamic programming are used to compute optimal defense policies, and heuristic look‑ahead policies are evaluated against optimal solutions. Throughout, the work emphasizes the importance of metrics (risk‑based rewards, defense‑mode costs, zone‑based reward structures) and illustrates how a managerial dashboard could integrate these metrics to support real‑time decision making. The dissertation concludes with a discussion of limitations, practical implications for IT managers, and a roadmap for future research, including extensions to richer attacker models, relaxed assumptions on influence matrices, and risk‑averse decision makers.
Topics
SecureRank: Risk‑Based Vulnerability Prioritization and Network‑Topology Effects
A quantitative scheme that ranks patches and security actions by their expected reduction in overall system risk. The model incorporates node interdependencies, vulnerability severity, and the influence of network topology (e.g., scale‑free, random, hierarchical) on risk propagation and patch effectiveness.