Resource-Sensitive Stealthy and Consensus-Kept State-Destructive Cyber Attacks for Multiagent Systems

Multiagent systems (MASs) are increasingly vulnerable to cyber attacks that exploit limited system resource to evade detection while causing disruptions. This paper investigates the resource-sensitive cyber attacks for MASs. Firstly, distributed replay attack (dRAs) and distributed pole-dynamics attack (dPDAs) are proposed, where dRAs utilize available resources of agents historical measurements, and dPDAs utilize available resources of agent parameter. Secondly, the relationship between available resources and stealthiness of cyber attack are revealed, where both dRAs and dPDAs can be with 0-stealthiness under different available resources. Thirdly, the relationship between available resources and destructiveness of cyber attack are revealed, where dRAs is not with ck-s-destructiveness (i.e., cyber attacks can achieve state trajectory destructiveness without disrupting consensus), because the state trajectory can only be driven to the historical one due to requiring historical measurements. Oppositely, dPDAs can be with ck-s-destructiveness, where the state consistency trajectory can be driven to any one designed by the attacker due to requiring agent parameter. Finally, numerical examples are provided to verify the effectiveness of the proposed attacks.