Hunt Kyle, Zhuang Jun
Department of Management Science and Systems, University at Buffalo, United States.
Department of Industrial and Systems Engineering, University at Buffalo, United States.
Eur J Oper Res. 2024 Mar;313(2):401-417. doi: 10.1016/j.ejor.2023.04.009.
In this article, we review the literature which proposes attacker-defender games to protect against strategic adversarial threats. More specifically, we follow the systematic literature review methodology to collect and review 127 journal articles that have been published over the past 15 years. We start by briefly discussing the common application areas that are addressed in the literature, although our focus in this review lies heavily in the approaches that have been adopted to model and solve attacker-defender games. In studying these approaches, we begin by analyzing the following features of the proposed game formulations: the sequence of moves, number of players, nature of decision variables and objective functions, and time horizons. We then analyze the common assumptions of perfect rationality, risk neutrality, and complete information that are enforced within the majority of the articles, and report on state-of-the-art research which has begun relaxing these assumptions. We find that relaxing these assumptions presents further challenges, such as enforcing new assumptions regarding how uncertainties are modeled, and issues with intractability when models are reformulated to account for considerations such as risk preferences. Finally, we examine the methods that have been adopted to solve attacker-defender games. We find that the majority of the articles obtain closed-form solutions to their models, while there are also many articles that developed novel solution algorithms and heuristics. Upon synthesizing and analyzing the literature, we expose open questions in the field, and present promising future research directions that can advance current knowledge.
在本文中，我们回顾了提出攻防博弈以防范战略对抗威胁的文献。更具体地说，我们遵循系统的文献综述方法，收集并回顾了过去15年发表的127篇期刊文章。我们首先简要讨论文献中涉及的常见应用领域，尽管在本综述中我们主要关注用于建模和解决攻防博弈的方法。在研究这些方法时，我们首先分析所提出的博弈公式的以下特征：行动顺序、参与者数量、决策变量和目标函数的性质以及时间范围。然后，我们分析大多数文章中所采用的完全理性、风险中性和完全信息的常见假设，并报告已开始放宽这些假设的前沿研究。我们发现放宽这些假设带来了进一步的挑战，例如在对不确定性进行建模时需要强化新的假设，以及在重新构建模型以考虑风险偏好等因素时出现的难处理问题。最后，我们研究了用于解决攻防博弈的方法。我们发现大多数文章都为其模型获得了闭式解，同时也有许多文章开发了新颖的求解算法和启发式方法。在综合和分析文献后，我们揭示了该领域的开放性问题，并提出了可以推动当前知识发展的有前景的未来研究方向。
