Abstract:Detecting and characterizing malicious operations in electricity markets presents substantial challenges due to the innate complexity of power networks. This paper explores adversarial attacks in electricity markets, i.e., we consider the scenario in which malicious market participants make undetectable changes to the input data of other participants, resulting in significant negative impacts on their operations. We propose a novel methodology to generate adversarial attacks, which leverages linear optimization theory and statistical learning to generate suboptimal, feasible solutions or super-optimal, infeasible solutions that convincingly resemble optimal solutions. The methodology is general and applicable for generating adversarial attacks on optimization-based decision systems. We use it to analyze the potential economic losses, policy making, and social consequences resulting from compromised demand-side operations. In particular, we focus on scenarios where certain demand side-players, i.e., microgrids, or prosumers in a general sense, engage in malicious activities by targeting other microgrids and the independent system operator. We model the electricity market dynamics using an equilibrium problem with equilibrium constraints (EPEC). To make the model easier to solve, the EPEC is reformulated using a combination of the Fisher–Burmeister function and strong duality theorem. Results show that microgrids can artificially decrease its consumption by up to 12% without triggering alarms. Additionally, this amount can be increased by an additional 14% to 22%, contingent on the strategy employed to monitor the observable variables. We also study the implications of the increasing utilization of solar generation, showing its potential to mitigate the impact of adversarial attacks while also introducing new risks to the system.

Adversarial Examples for Model-Based Control: A Sensitivity Analysis

Highly transferable adversarial attack against deep‐reinforcement‐learning‐based frequency control

Stealthy Actuator Signal Attacks in Stochastic Control Systems: Performance and Limitations.

Learning-Based Resilient FCS-MPC for Power Converters under Actuator FDI Attacks

Adversarial Model Predictive Control via Second-Order Cone Programming

Controller Synthesis for Linear Time-varying Systems with Adversaries

Tube Model Predictive Control Based Cyber-Attack-Resilient Optimal Voltage Control Strategy in Wind Farms

Adversarial attacks in demand-side electricity markets

Optimal Attack against Autoregressive Models by Manipulating the Environment

Resilient Feedback Controller Design For Linear Model of Power Grids

A Practical Adversarial Attack on Contingency Detection of Smart Energy Systems

Performance-Robustness Tradeoffs in Adversarially Robust Control and Estimation

A Novel Bifurcation Method for Observation Perturbation Attacks on Reinforcement Learning Agents: Load Altering Attacks on a Cyber Physical Power System

An Efficient Resilient MPC Scheme via Constraint Tightening against Cyberattacks: Application to Vehicle Cruise Control

Adversarial Robustness for Tabular Data through Cost and Utility Awareness

A black-box adversarial attack on demand side management

Exploring Adversarial Threat Models in Cyber Physical Battery Systems

Generating Adversarial Disturbances for Controller Verification

An Adversarial Approach to Evaluating the Robustness of Event Identification Models

Synthesis-guided Adversarial Scenario Generation for Gray-box Feedback Control Systems with Sensing Imperfections

Online Optimization of LTI Systems Under Persistent Attacks: Stability, Tracking, and Robustness