An extended study on multi-objective security games

Abstract

The burgeoning area of security games has focused on real-world domains where security agencies protect critical infrastructure from a diverse set of adaptive adversaries. In such domains, decision makers have multiple competing objectives they must consider which may take different forms that are not readily comparable including safety, cost, and public perception. Thus, it can be difficult to know how to weigh the different objectives when deciding on a security strategy. To address the challenges of these domains, we propose a fundamentally different solution concept, multi-objective security games (MOSG). Instead of a single optimal solution, MOSGs have a set of Pareto optimal (non-dominated) solutions referred to as the Pareto frontier, which can be generated by solving a sequence of constrained single-objective optimization problems (CSOP). The Pareto frontier allows the decision maker to analyze the tradeoffs that exist between the multiple objectives. Our contributions include: (i) an algorithm, Iterative--Constraints, for generating the sequence of CSOPs; (ii) an exact approach for solving an MILP formulation of a CSOP; (iii) heuristics that achieve speed up by exploiting the structure of security games to further constrain the MILP; (iv) an approximate approach for solving a CSOP built off those same heuristics, increasing the scalability of our approach with quality guarantees. Additional contributions of this paper include proofs on the level of approximation, detailed experimental evaluation of the proposed approaches and heuristics, as well as a discussion on techniques for visualizing the Pareto frontier.

Security Games in the Field: Deployments on a Transit System

“…In contrast, MOPSS is the first deployed system to use three significantly different adversary models to develop three different patrol schedules for the threats of fare evasion, terrorism and crime. In contrast with previous work suggesting such threats could be modeled as a multi-objective security game [4], A fundamental contribution of this paper is the insight that these different threat types lead to fundamentally different adversary models that cannot be folded into a single security game framework. MOPSS then is built upon these three adversary models.…”

“…Our first contribution is MOPSS, the first Multi-Operation Patrol Scheduling System for patrolling a train line. MOPSS provides an important insight: the multiple threats (FE, CT and CR) in a transit system require such fundamentally different adversary models that they do not fit into state-of-the-art multiobjective or Bayesian security game models suggested earlier [18,4]. Instead, in MOPSS each of the three threats is modeled as a separate game with its own adversary model.…”

Derivative-Free Reinforcement Learning: A Review

“…Therefore, derivative-free optimization methods are suitable for dealing with the complex and hard optimization problems. They have been applied in the complex learning tasks and achieved impressive empirical results, such as policy search in reinforcement learning [Taylor et al, 2006;Abdolmaleki et al, 2015;Salimans et al, 2017], automatic machine learning and hyper-parameter tuning [Snoek et al, 2012;Thornton et al, 2013;Real et al, 2017Real et al, , 2018, objective detection in computer vision [Zhang et al, 2015], subset selection , and security games [Brown et al, 2014], etc.…”

Hiding Signal Strength Interference from Outside Adversaries

“…In contrast our work provides a rigorous framework for characterizing when a person might or statistically cannot be detected by any method, or bounds the rate of potential detection up to some statistical confidence. A different tact to characterizing when detection is or is not possible uses game-theoretic approaches [9], [16], [21], as opposed to our statistical/information theory approach.…”