| 摘要: |
| 智能电网、大型工业系统等国家关键基础设施是大规模人机物融合网络,安全威胁来自信息域、物理域和社会域,系统安全性的实现需要多域综合的智能对抗。本文针对攻击者侦察、入侵和破坏等环节的高级威胁,研究人机物融合的面向可观性与可控性的智能对抗模型和新机制。首先基于关键基础设施对物理子系统的感知与控制功能来刻画人机物融合模型,然后描述宏观的复杂网络模型和微观的实体关系模型;刻画的安全威胁模型描述了攻击者如何利用人机物的关联特性和脆弱性,来实施跨域渗透攻击与多域协同攻击。为应对攻击者侦察阶段的多域目标探测和跨域渗透,引入人机物多域协同的移动目标防御,提出多层协同的动态反渗透博弈对抗模型;为应对多域协同入侵行为的隐蔽性,设计了人机物多通路完整性监测框架,引入人机物多域关联的异常检测方法;为应对多域协同攻击,设计信息物理联动的安全机制和信息安全感知的控制算法,通过对操作员行为不确定性的调控和认知非理性的修正实现人机物多域联动的安全增强。我们提出完整的以人为中心的安全对抗模型,通过引入社会域以人为中心的移动目标防御,减少人作为攻击入口的风险;通过引入以人为中心的网络行为监测,实现环路观人的监测结构;通过消减人在认知过程中感知、注意、记忆、学习、决策等环节的脆弱性,来增强人在环路的安全控制能力。本文提出的模型将有助于丰富网络空间安全基础模型和发展人机物融合的计算范式,增强关键基础设施安全性。 |
| 关键词: 关键基础设施 人机物融合 移动目标防御 入侵检测 安全博弈 |
| DOI：10.19363/J.cnki.cn10-1380/tn.2025.07.09 |
| Received:November 03, 2023Revised:January 21, 2024 |
| 基金项目:本课题得到国家自然科学基金(No. 61572514),湖南省自然科学基金(No. 2023JJ30085, No. 2023JJ40272),湖南省教育厅项目(No.22A0599, No. 20A511, No. 22B0938)资助。 |
|
| Human-Cyber-Physical Collaborative Countermeasure Models for Critical Infrastructure Security |
| ZHU Peidong,KANG Wenjie,LIU Liang,ZHANG Rui,XUN Peng |
| School of Electronic Information and Electrical Engineering, Changsha University, Changsha 410022, China;Information technology Department, Hunan Police Academy, Changsha 410138, China;College of Computer, National University of Defense Technology, Changsha 410073, China |
| Abstract: |
| Critical infrastructures such as smart grids and large industrial systems are large-scale human-cyber-physical networks, with security threats coming from the cyber domain, physical domain, and social domain. The implementation of system safety requires comprehensive intelligent countermeasures from multiple domains. This paper focuses on advanced threats throughout attacker reconnaissance, invasion, and destruction, and studies intelligent human-cyber-physical countermeasure models and new mechanisms for observability and controllability. We first characterize the human-cyber-physical model based on the monitoring and control functions of critical infrastructure on physical subsystems, and then model it from the perspectives of macro complex networks and micro entity relationships; the proposed security threat model describes how attackers can fully utilize the correlation and vulnerability of human, cyber and physical elements to implement cross-domain infiltration attacks and multi-domain collaborative attack. To cope with the multi-domain target exploration and cross-domain penetration, a multi-domain collaborative mobile target defense model is introduced, and a multi-layer collaborative dynamic game model against penetration is proposed; to address the stealth of multi-domain collaborative intrusion behavior, a human-cyber-physical multi-pathway framework for integrity monitoring is designed, and an anomaly detection method based on human-cyber-physical multi-domain correlation is introduced; to deal with multi-domain collaborative attacks, security mechanisms for cyber-physical collaboration and security-aware control algorithms are to be designed, and by regulating the uncertainty of operator behavior and correcting cognitive irrationality, the security enhancement of human-cyber-physical coordination is achieved. The proposed human-centered security countermeasure model is relatively complete, which reduces the risk of humans as attack entry points by introducing human-centered mobile target defense in the social domain; by introducing a human-centered network behavior monitoring approach, a monitoring structure for observing people on the loop is achieved; by reducing the fragility of perception, attention, memory, learning, decision-making in the cognitive process, we can enhance our safety control ability in the loop. These models will help strengthen the fundamental framework of cybersecurity and develop novel paradigms for human-cyber-physical computation, and boost the security of critical infrastructure. |
| Key words: critical infrastructure human-cyber-physical convergence mobile target defense intrusion detection security game |
