Distribution System Reconfiguration to Mitigate Load Altering Attacks via Stackelberg Games

TL;DR

This paper develops a game-theoretic approach using distribution network reconfiguration to defend against load-altering cyberattacks on IoT devices, ensuring system safety and robustness.

Contribution

It introduces a novel Stackelberg game framework with Bayesian optimization for efficient, strategic defense against load-altering attacks in power distribution systems.

Findings

The proposed method effectively reduces voltage violations caused by attacks.

The Stackelberg game approach enhances system resilience against strategic adversaries.

Bayesian optimization significantly speeds up equilibrium computation.

Abstract

The widespread integration of IoT-controllable devices (e.g., smart EV charging stations and heat pumps) into modern power systems enhances capabilities but introduces critical cybersecurity risks. Specifically, these devices are susceptible to load-altering attacks (LAAs) that can compromise power system safety. This paper quantifies the impact of LAAs on nodal voltage constraint violations in distribution networks (DNs). We first present closed-form expressions to analytically characterize LAA effects and quantify the minimum number of compromised devices for a successful LAA. Based on these insights, we propose a reactive defense mechanism that mitigates LAAs through DN reconfiguration. To address strategic adversaries, we then formulate defense strategies using a non-cooperative sequential game, which models the knowledgeable and strategic attacker, accounting for the worst-case scenario and enabling the reactive defender to devise an efficient and robust defense. Further, our formulation also accounts for uncertainties in attack localization. A novel Bayesian optimization approach is introduced to compute the Stackelberg equilibrium, significantly reducing computational burden efficiently. The game-theoretic strategy effectively mitigates the attack's impact while ensuring minimal system reconfiguration.