A Bayesian computational methodology to solve general security games
Security games provide a robust and flexible framework for modeling strategic and operational challenges in defense and homeland security (DHS). However, these games are often studied using game-theoretic approaches that rely on the assumption of common knowledge—an assumption that rarely holds in DHS contexts. Adversarial Risk Analysis (ARA) offers a Bayesian alternative that addresses these limitations.
In this talk, we introduce a computational methodology for analyzing general security games from an ARA perspective. Our approach is capable of resolving any game represented as a bi-agent Influence Diagram, accommodating both discrete and continuous decision domains and supporting an arbitrary number of decisions per agent. We also discuss the computational challenges posed by continuous decision spaces and multi-stage decision processes. To illustrate the practical applicability of our methodology, we present a case study on disinformation warfare.
Keywords: Adversarial risk analysis Augmented probability simulation Security games Disinformation