[Paper Review] A Survey of Graph-Theoretic Approaches for Analyzing the Resilience of Networked Control Systems

Mohammad Pirani, Aritra Mitra+1 more|arXiv (Cornell University)|May 25, 2022

Smart Grid Security and Resilience4 citations

TL;DR

This survey presents graph-theoretic frameworks to analyze and enhance the resilience of networked control systems (NCS) against adversarial attacks. By leveraging network structure properties such as robustness, isoperimetric constants, and controllability Gramian spectra, the paper reinterprets system-theoretic robustness concepts and identifies key design principles for secure, distributed algorithms with minimal communication and maximal fault tolerance under targeted attacks.

ABSTRACT

As the scale of networked control systems increases and interactions between different subsystems become more sophisticated, questions of the resilience of such networks increase in importance. The need to redefine classical system and control-theoretic notions using the language of graphs has recently started to gain attention as a fertile and important area of research. This paper presents an overview of graph-theoretic methods for analyzing the resilience of networked control systems. We discuss various distributed algorithms operating on networked systems and investigate their resilience against adversarial actions by looking at the structural properties of their underlying networks. We present graph-theoretic methods to quantify the attack impact, and reinterpret some system-theoretic notions of robustness from a graph-theoretic standpoint to mitigate the impact of the attacks. Moreover, we discuss miscellaneous problems in the security of networked control systems which use graph-theory as a tool in their analyses. We conclude by introducing some avenues for further research in this field.

Motivation & Objective

• To address the growing challenge of securing large-scale, distributed NCSs against intelligent, targeted adversarial attacks rather than random faults.
• To bridge systems and control theory with graph theory by reinterpreting robustness and resilience using structural network properties.
• To identify minimal communication requirements for achieving desired resilience levels in distributed algorithms.
• To explore new research directions such as nonlinear interactions, multi-agent reinforcement learning, and attack energy minimization.

Proposed method

• Uses graph-theoretic measures like network robustness, isoperimetric constants (i(G)), and algebraic connectivity (λ₂(L)) to quantify resilience.
• Applies the concept of r-robustness to ensure distributed algorithms remain functional under adversarial node removal or misinformation.
• Analyzes attack impact through the spectra of the controllability Gramian W_F to model attack energy and system vulnerability.
• Reinterprets classical control notions—like observability and controllability—using graph structure to guide resilient algorithm design.
• Examines point-to-point communication vulnerabilities where attackers can inject inconsistent data, unlike consistent fault propagation.
• Proposes minimizing edges while maintaining resilience, linking network design to controllability and robustness constraints.

Experimental results

Research questions

• RQ1 How can graph-theoretic properties such as isoperimetric constants and r-robustness be used to quantify the resilience of distributed algorithms in NCSs?
• RQ2 What is the relationship between algebraic connectivity (λ₂(L)) and network robustness, and how can it guide the design of resilient networks?
• RQ3 How can the energy of an adversarial attack be modeled and minimized using the spectra of the controllability Gramian?
• RQ4 What are the structural vulnerabilities of scale-free and random networks in the context of targeted node removal or misinformation?
• RQ5 How can resilience be maintained in NCSs with nonlinear dynamics, such as Kuramoto oscillators or swarm robotics systems?

Key findings

• If the isoperimetric constant i(G) > r−1, then the network is at least r-robust, providing a structural condition for resilience against adversarial attacks.
• A lower bound on network robustness is given by ⌊r/2⌋ when λ₂(L) > r−1, though this bound is often loose, as shown by the star graph example with λ₂(L)=1 and 1-robustness.
• The controllability Gramian's spectrum provides a tool to quantify attack energy, suggesting that network design can be optimized to maximize this energy and thus increase attack difficulty.
• Minimal communication networks can be designed to maintain resilience, but minimizing edges while preserving network robustness remains an open problem.
• Scale-free networks are highly vulnerable to targeted hub-node attacks, which can disconnect the network with minimal effort, highlighting the need for structural hardening.
• Extending current graph-theoretic resilience methods to nonlinear systems (e.g., Kuramoto oscillators) and multi-agent reinforcement learning remains an open and critical research frontier.