Homeland security resource allocation games: Considering partially strategic attackers and equity

Introduction The terrorist attacks of September 11, 2001, have drawn huge amounts of money into homeland security in the United States. For instance, Figure 27.1(a) shows the annual budget requests for the U.S. Department of Homeland Security (DHS) since it was formed in 2002. However, the effectiveness of such large expenditures is unknown. “Pork-barrel politics,” where political factors are driving resources toward low-risk targets, is used by critics of how homeland security resources are utilized (e.g., de Rugy, 2005). Toward risk-based optimization of resource allocation, DHS has employed a three-step method, which is based on perceived risk, in directing grant distribution since 2006 (U.S. Government Accountability Office, 2008). Construction of a comprehensive risk management framework would mandate more theoretical and practical work. Operations research (OR) techniques have proven valuable in developing counterterrorism strategies. Harris (2004) highlighted how probabilistic risk analysis, randomization, and game theory could play a role in defending against attacks initiated by terrorists. Brown and colleagues (2005) built a two-sided optimization model whose goal is to optimally pre-localize defensive missile platforms in the face of adaptive attackers. By modeling and data analysis, Kaplan and Kress (2005) studied the effectiveness of a suicide-bomber-detector scheme in decreasing fatalities resulted from suicide bombing. Brown and colleagues (2006) constructed innovative bi-level and tri-level models to explore optimal strategies to defending critical infrastructure. Lin and colleagues (2009) studied a M/G/1 queue to develop optimal scheduling policies to be effectively used by an antiterrorist surveillance system. Wein (2009) discussed how mathematical modeling can assist with policy recommendations. With queuing theory and Markov population processes, Kaplan (2010) investigated the optimal strategies to be employed by undercover intelligence agents in order to infiltrate and stop terror plots. Jose and Zhuang (2013) considered technologies adoption of an attacker, which is repeatedly interacting with the same defender using sequential games with uncertainty. One main contribution of OR methods to tackling homeland-security problems is to understand the attacker's objectives. This could be achieved by building “value models” (Keeney, 2007). For instance, Wang and Bier (2010, 2011) built a multiattribute utility function to model terrorist's preference behaviors. Different approaches of weighting multiple attributes have been compared by Stillwell and colleagues (1987).