Validating Models of Adversary Behavior

Abstract:  The objectives of this project are: (a) to integrate and stimulate research in the area of model validation of adaptive adversaries; (b) to transform existing models of adversary behavior and demonstrate their validity and applicability to real-world problems; (c) to transit the results to homeland security practitioners; and (d) to provide education and outreach to a large group of governmental agencies, students and communities.   This project has been developing and transitioning models in integrating game theory and operations research for both terrorism and natural disasters, resulting in numerous peer-reviewed journal publications conference/seminar presentations, education and outreach, and internal/external awards. The research has interested agencies from DHS and other government organizations. The PI is successfully organizing the Second conference on Validating Models of Adversary Behavior, Buffalo/Niagara Falls, NY, August 2-5, 2015, which attracted 20 technical talks, 3 poster presentations, and ~30 attendees (including some CREATE researchers) in the field from around the world. Although one of the consensuses is that model validation in adversary modeling is challenging and often not feasible due to the lack of publicly available data for rare events. However, many alternatives and methods are being presented and discussed. The PI is editing a special issue of Risk Analysis on Validating Models of Adversary Behavior, documenting the results of the first conference. The PI co-edited two Springer books Game Theoretic Analysis of Congestion, Safety and Security: Networks, Air Traffic and Emergency Departments (ISBN 978-3-319-13009-5) and Game Theoretic Analysis of Congestion, Safety and Security: Traffic and Transportation Theory (ISBN 978-3-319-11674-7).   The research is applicable to DHS and other government agencies, especially in homeland security resource allocation in the face of adaptive adversary. The theoretical models have matured in the past years to capture more realistic considerations, including more players (local/federal/foreign governments, adversaries, private citizens/communities), more player options (e.g., attack/defense/contract/subsidy), more attributes in objective functions (e.g., balancing between security and congestion/waiting, balancing between equity and efficiency), and more realistic adversary behavior (e.g., partially strategic). All model parameters are tunable so that practitioners are able to customize and play with the software demos.   Through simulation and experiments, this project integrated and stimulated research in the area of model validation, a research domain that is underdeveloped, especially in the area of counter terrorism. By bridging theoretical and empirical bodies of research in adversarial modeling, this project has helped transform existing models of adversary behavior and demonstrate their validity and applicability to real-world homeland security problems. This project will help identify and quantify the accuracy of models in predicting reality, which is an under-studied area.   This project has created important synergies of communication between researchers from CREATE (primarily risk analysts, quantitative economists, and game theorists) and National Consortium for the Study of Terrorism and Responses to Terrorism (START, primarily social scientists). For example, the PI has been working with three START investigators on other projects and proposals, related to this project. In the long run, this project will help transition models of adversary behavior and demonstrate their validity and applicability to real-world problems by bridging theoretical and empirical bodies of research in adversarial modeling.   This has been again a productive year for this CREATE-sponsored project. We have 11 peer-reviewed journal papers published/accepted. The research group (the PI and his students) made 12 conference presentations and the PI spoke at over 14 invited research seminars and guest lectures. The PI graduated two Ph.D. students and a couple of MS students with thesis. The PI received the 2014 MOR Journal Award, for the best paper published in 2013 in the journal Military Operations Research, jointly awarded by the INFORMS’ Military Applications Society and the Military Operations Research Society.