Attack Analysis for Distributed Control Systems: An Internal Model Principle Approach
For researchers in distributed control systems, this work provides a system-theoretic understanding of worst-case attack effects, highlighting the need for more resilient control protocols.
This paper analyzes cyber-physical attacks in distributed multi-agent systems, showing that a stealthy attack on root nodes can mislead the entire network and destabilize synchronization, formalized as an internal model principle for the attacker.
Although adverse effects of attacks have been acknowledged in many cyber-physical systems, there is no system-theoretic comprehension of how a compromised agent can leverage communication capabilities to maximize the damage in distributed multi-agent systems. A rigorous analysis of cyber-physical attacks enables us to increase the system awareness against attacks and design more resilient control protocols. To this end, we will take the role of the attacker to identify the worst effects of attacks on root nodes and non-root nodes in a distributed control system. More specifically, we show that a stealthy attack on root nodes can mislead the entire network to a wrong understanding of the situation and even destabilize the synchronization process. This will be called the internal model principle for the attacker and will intensify the urgency of designing novel control protocols to mitigate these types of attacks.