Anomaly Detection in Encrypted Inverted Pendulum Systems: Simulation-based Parameter Analysis
Sangwon Lee, Junsoo Kim · Journal of Institute of Control Robotics and Systems · 2025
In networked control systems using homomorphic encryption, detecting anomalies without decrypting the encrypted data remains a challenge. Prior studies have proposed methods that use zero dynamics to disclose residual signals by modifying the learning with errors (LWE)-based encryption scheme, enabling anomaly detection directly on encrypted data while maintaining security. In this paper, we apply this approach to an inverted pendulum system and implement an encrypted controller through simulation to validate anomaly detection. We then choose appropriate encryption and quantization parameters that satisfy the performance requirements of the encrypted controller under given constraints. Finally, we compare the theoretical upper bounds of the quantization error with simulation results with respect to parameter choices. Simulation results demonstrate that anomaly detection can be performed on encrypted data using the proposed approach, while also confirming the feasibility of the proposed parameter settings for potential implementation in real systems.