Improving Homeland Security with Decision Analysis

Principal Investigator: Detlof von Winterfeldt


Year 9 marked the conclusion of two project tasks, which had been conducted in Years 7-9 with funding from CREATE:Attacker-Defender 

  1. Decision Tree Analysis
  2. Development of a DHS Utility Function

The purpose of the first task was to develop and apply attacker-decision tree models to important problems faced by the DHS.  Defender-attacker decision trees are closely related to leader-follower (von Stackelberg) games, which were adapted for the terrorism context by Milind Tambe and his colleagues at CREATE.  Instead of optimizing the defender-attacker decisions at each stage of the extended game, attacker behavior is modeled probabilistically, conditional on defender decisions.  This approach was previously applied in a two-stage and multi-stage decision tree for the MANPADS problem and in a two-stage decision tree for choosing technologies and configurations for radiation detection portals.  In year 9 we developed a framework for applying attacker-defender decision trees to the multi-layered defenses of the Global Nuclear Detection Architecture (GNDA).  This work was done in collaboration with the National Academy Committee “Assessing the Effectiveness of the Global Nuclear Defense Architecture,” which conducted its work in 2012 and 2013. An important part of this work is to understand the attacker’s objectives.  In the past, we have identified and structured objectives of Al Quaeda, using their published statements and website materials (see Keeney and von Winterfeldt, 2010). During Years 8 and 9, we used the same methodology to identify and structure the objectives of Hezbollah (Behruz, Rosoff and von Winterfeldt, 2013).   The main purpose of the second task was to develop a DHS utility function.  To achieve this, we developed a comprehensive set of DHS objectives, associated metrics and tradeoffs among the objectives.  This was a preliminary exercise that was based on a survey and a literature review. Tradeoffs were developed primarily on the basis of a literature review.  Future work will extend this work with interviews with policy makers. In Year 9, we applied this framework to the GNDA, resulting in a set of notional objectives and metrics that were presented to the DNDO leadership.