VIMU: Effective Physics-based Realtime Detection and Recovery against Stealthy Attacks on UAVs

TL;DR

VIMU addresses sensor-attacks on UAVs by integrating a high-fidelity nonlinear physical model with a fast, robust anomaly detector (CS-EMA) and a FIFO-buffer safeguard to preserve flight stability even when all IMUs are compromised. The framework comprises four modules—anomaly detectors, a recovery monitor, a state estimator, and a physical model—working together to detect, isolate, and recover from attacks, while preserving continuous control through a reference state when sensors fail. Key contributions include the CS-EMA detector, a fine-grained aerodynamic/propulsion model, lattice-based recovery policies, and an efficient PX4 implementation with demonstrated improvements in detection latency and recovery duration over prior PBAD/PBAR approaches. Real-world and simulator experiments show that VIMU achieves faster detection (as low as tens of milliseconds) and longer recovery windows, enabling flight stability under stealthy and overt sensor attacks and making it practical for deployment on resource-constrained UAV platforms.

Abstract

Sensor attacks on robotic vehicles have become pervasive and manipulative. Their latest advancements exploit sensor and detector characteristics to bypass detection. Recent security efforts have leveraged the physics-based model to detect or mitigate sensor attacks. However, these approaches are only resilient to a few sensor attacks and still need improvement in detection effectiveness. We present VIMU, an efficient sensor attack detection and resilience system for unmanned aerial vehicles. We propose a detection algorithm, CS-EMA, that leverages low-pass filtering to identify stealthy gyroscope attacks while achieving an overall effective sensor attack detection. We develop a fine-grained nonlinear physical model with precise aerodynamic and propulsion wrench modeling. We also augment the state estimation with a FIFO buffer safeguard to mitigate the impact of high-rate IMU attacks. The proposed physical model and buffer safeguard provide an effective system state recovery toward maintaining flight stability. We implement VIMU on PX4 autopilot. The evaluation results demonstrate the effectiveness of VIMU in detecting and mitigating various realistic sensor attacks, especially stealthy attacks.

Figures (23)

• Figure 1: High-level Overview of VIMU Modules
• Figure 2: Workflow of Anomaly Detection
• Figure 3: Lattice-based Recovery Policy for Flight Altitude ($p_z$) and Angular Velocity ($\omega$). "_" Stands for Any Possible Cases and "$\bot$" Stands for Sensor Instances are Compromised
• Figure 4: Workflow of Front-End EKF of VIMU-SE
• Figure 5: TPR Heatmap on Overt Gyro Attacks