计算机科学
数字水印
分离(微生物学)
乘法函数
分布式计算
选择(遗传算法)
多智能体系统
计算机网络
实时计算
模型攻击
信号(编程语言)
计算机安全
计算复杂性理论
联轴节(管道)
作者
Changda Zhang,Jingbo Wang,Zhijie Li,Dajun Du,Changchun Hua
标识
DOI:10.1109/jiot.2026.3658079
摘要
The communication links of multiagent systems (MASs) could be compromised by stealthy cyber attacks. To solve the problem, this paper proposes distributed nodes detection and event-triggered links isolation of stealthy attacks in MASs by topology-oriented watermarking (TOW). Firstly, the limitation of traditional MASs against distributed pole-dynamics attacks (dPDAs) is revealed. Secondly, different from existing water-marking methods for single-agent systems, a novel TOW method with four optional parameters selection schemes is proposed for distributed nodes detection in MASs, where it is proved that only one of optional parameters selection schemes can simultaneously possesses complexity and resistance to collusive dPDAs. Moreover, TOW combines the advantages of multiplicative and additive watermarking to achieve dPDAs detection. Furthermore, the positive relationship between dPDAs detection performance and the coupling factor of attack topology, attack signal and TOW parameters is quantified. Thirdly, a new TOW-based event-triggered links isolation algorithm incorporating multiple link-deletion observers is proposed, where TOW is leveraged to trigger running of link-deletion observers. It is proved that the compromised links can be correctly isolated. Finally, the simulation results verify our proposed scheme.
科研通智能强力驱动
Strongly Powered by AbleSci AI