基于非确定性概率型信息的智能安全决策系统研究

公共安全乃国之大计，如何利用人工智能技术去预防和阻止实时犯罪行为是社会安全管理的发展热点。传统智能安全系统主要关注于理想实验室环境下的潜在威胁识别，较少涉及现实非确定性概率型信息下的多个潜在威胁排序及有限资源下的潜在威胁排除等相关的安全决策研究，因此其在突发危机事件中，并不能为相关的安全人员提供很好的决策支持和建议。针对该问题，本项目将利用不确定决策理论、复杂事件刻画与推理系统和安全博弈的相关知识，构建基于非确定性概率型信息的智能安全决策系统。具体研究内容叙述涵括三个方面：（1）构建非确定性概率型信息下的公理化决策模型；（2）建立实时安全博弈模型并给出最优解概念，从理论上证明智能安全决策系统的可行性；（3）结合实时安全博弈模型和复杂事件刻画和推理系统，构建非确定性概率型信息下的智能安全决策系统，实现威胁识别，威胁危害性排序，多威胁行为下资源数量分配，实时安全资源分配等多目标整合。

Public security is a critical concern of our country. How to use artificial intelligence technology to prevent and stop real-time crime is the hot spots of social security management. The traditional intelligent security systems mainly focus on the identification of potential threats under ideal laboratory environment. However, in real-world applications, few existing system can directly solve the security decision making problem, such as the ranking of potential threat and threat prevention based on limited security resource under ambiguity information. Therefore, in the case of sudden crisis, none of them can provide decision support or decision advice for the security manager in real world applications. In order to address the issue, in this project, based on the decision theory under ambiguity, complex event modeling and reasoning system, and security game, we will construct an intelligence security decision making system based on the ambiguity information. More specifically, we will first construct a decision model under ambiguity. And then, we will construct a real-time security game model and find out the optimal solution based on the ambiguity information. After that, we will combine the real-time security game model and complex event modeling and reasoning system to build the intelligence security decision making system. Therefore, we can make the threat identification, ranking of potential threat, the security resource allocation for multiple threats, and so on.