Event-Triggered Secure Control Design Against False Data Injection Attacks via Lyapunov-Based Neural Networks

Abstract

This paper presents a secure control framework enhanced with an event-triggered mechanism to ensure resilient and resource-efficient operation under false data injection (FDI) attacks on sensor measurements. The proposed method integrates a Kalman filter and a neural network (NN) to construct a hybrid observer capable of detecting and compensating for malicious anomalies in sensor measurements in real time. Lyapunov-based update laws are developed for the neural network weights to ensure closed-loop system stability. To efficiently manage system resources and minimize unnecessary control actions, an event-triggered control (ETC) strategy is incorporated, updating the control input only when a predefined triggering condition is violated. A Lyapunov-based stability analysis is conducted, and linear matrix inequality (LMI) conditions are formulated to guarantee the boundedness of estimation and system errors, as well as to determine the triggering threshold used in the event-triggered mechanism. Simulation studies on a two-degree-of-freedom (2-DOF) robot manipulator validate the effectiveness of the proposed scheme in mitigating various FDI attack scenarios while reducing control redundancy and computational overhead. The results demonstrate the framework's suitability for secure and resource-aware control in safety-critical applications.