Collaborative Research: Network Protection under Uncertainty

本研究项目侧重于在恐怖威胁的情况下提高人口的安全性，以及如何使用集成建模(1)提高对对抗环境下人类决策的科学理解，以及(2)设计保护软目标（即公共交通设施、群众聚集场所及其用户）的分配政策和最佳实践。它涉及理论和计算研究，将通过综合建模和优化探索基础设施保护技术的基础，并将开发新的方法和工具，以支持实时应急管理和政策制定。该项目具有广泛的应用范围，从国土安全和执法到军事应用。这项研究具有变革性，因为它为风险管理和安全传感/控制网络带来了新的视角。该研究预计将对公共设施、校园、学校建筑及其用户产生最重大的影响，为第一响应者提供有关预防和威慑恐怖袭击的决策支持，以及在袭击期间最大限度地减少损失。工作结果将与主要的区域参与者进行沟通，包括纽约/新泽西港务局、新泽西运输局、新泽西国土安全部和新泽西交通部。该研究将产生的对抗性决策模型和虚拟游戏基础设施也适用于其他领域，如培训执法人员防止帮派暴力和培训校园警察。这项研究还力求产生广泛的影响，特别是在研究和教育的结合方面；它将为博士论文提供基础，并将为指导信息科学及其在安全领域的应用的博士后提供支持。研究结果将为基础设施保护游戏的综合书籍建立框架，供学生和从业者使用。该研究项目的主要特点是：(i)封闭形式的分析解决方案，（ii）探索玩家可以利用的关于对手类型、攻击类型和力量的不确定性和不完整信息的方式的游戏模型，（iii）包含检测概率和代表移动目标的人员流动模型的巡逻游戏，（iv）在沉浸式虚拟世界中构建虚拟游戏，旨在收集玩家行为数据，并使用预期伤害、失败攻击比例等指标验证游戏模型。研究的目标是：(1)开发随机网络保护博弈算法，供防御者使用，结合现有技术，规划应急准备，预防恐怖主义或网络攻击，以减少宝贵资产的损失；(2)设计一个包含上述模型关键要素的虚拟游戏，收集游戏中各类玩家的行为信息；(3)通过虚拟游戏在受控环境中测试真实玩家行为模型，以验证和完善随机游戏中人类受试者行为模型。这些目标将通过调整资源分配博弈模型来实现。将收集真实的玩家行为数据来验证模型。本研究的预期结果既具有分析性，又与从业者相关，如应急管理机构、警察等。该研究为人为突发事件中对抗政策的随机建模做出了贡献；部分可观测系统的随机博弈论；Iii)保卫关键基础设施的决策；Iv)在国土安全和资源分配方面教育未来的急救人员。

This research project focuses on improving the security of the population in the case of a terrorist threat, and how integrated modeling can be used (1) to improve scientific understanding of human decision-making in adversarial contexts, and (2) in the design of allocation policies and best practices for protection of soft targets, i.e., public transit facilities, places of mass gathering, and their users. It involves theoretical and computational research that will explore the foundations of infrastructure protection technologies through integrated modeling and optimization and will develop new methods and tools to support real-time emergency management and policy development. The project has a wide range of applications from homeland security and law enforcement to military applications. This research is transformational since it brings a fresh vision to risk management and sensing/control networks for security. The most significant impact of the research is expected to be to public facilities, campuses, school buildings and its users by providing support to first responders in decision making about prevention and deterrence of terrorist attacks as well as minimization of damage during an attack. Results of the work will be communicated with the major regional players, including NY/NJ Port Authority, NJ Transit, NJ Department of Homeland Security and NJDOT. The adversarial decision making models and virtual game infrastructure that will result from the research are also applicable to other areas such as training law enforcement personnel to prevent gang violence and training campus police. This study also seeks to have a broad impact, especially, in regard to the integration of research and education; it will provide the basis for a Ph.D. dissertation and will provide support to mentor a postdoctoral associate in information sciences and its application to security. The results of the research will establish the framework for a comprehensive book on infrastructure protection games to be used by students as well as practitioners.The key features of the research project are (i) closed form analytical solutions, (ii) game models that explore the ways in which the uncertainty and incomplete information about the adversary type, attack type and power, can be exploited by the players, (iii) a patrolling game that incorporates detection probability and a people flow model representing moving targets, and (iv) a virtual game built inside an immersive virtual world designed to both collect data about the players' behavior and to validate the game models using metrics such as the expected damage, the fraction of unsuccessful attacks, etc. The objectives of the research are to (1) Develop stochastic network protection game algorithms to be used by the defenders, together with current technologies, in planning for emergency preparedness, for prevention of terrorist or cyber attacks to reduce the loss of valuable assets; (2) Design a virtual game involving the key elements of the above models in order to collect information about the behavior of various players of the game; (3) Test models for real player behavior in a controlled setting via virtual games to verify and refine models for human subject behavior in the stochastic games. These objectives will be accomplished by adapting resource allocation game models to the virtual game. Realistic player behavior data will be collected to validate the models. The anticipated results of this research are both analytical and relevant to practitioners, such as emergency management agencies, police, etc. The research makes contributions to i) stochastic modeling of adversarial policies in man-made emergencies; ii) stochastic game theory for partially observable systems; iii) decision-making in defending critical infrastructure; iv) educating future first responders in homeland security and resource allocation.