MARS: Defending Unmanned Aerial Vehicles From Attacks on Inertial Sensors with Model-based Anomaly Detection and Recovery

TL;DR

IMU attacks on UAVs can destabilize attitude control and cause crashes; the paper introduces MARS, a model-based anomaly detection and recovery framework that bypasses compromised IMUs by fusing tachometer rotor-speed data with heading and position sensors through an EKF-based resilient state estimator. A dedicated anomaly detector (CUSUM with sliding window) provides fast system-level alerts, triggering a multi-stage recovery that first brakes to near-hover and then resumes the mission under resilient control. The authors implement MARS on PX4 (MARS-PX4), validate in software-in-the-loop Prometheus/Gazebo and on real-world X500 platforms, and show superior survival time and smoother control compared to existing IMU-defense methods. The results demonstrate practical UAV resilience to both acoustic resonant and EMI attacks, enabling continued operation under attack.

Abstract

Unmanned Aerial Vehicles (UAVs) rely on measurements from Inertial Measurement Units (IMUs) to maintain stable flight. However, IMUs are susceptible to physical attacks, including acoustic resonant and electromagnetic interference attacks, resulting in immediate UAV crashes. Consequently, we introduce a Model-based Anomaly detection and Recovery System (MARS) that enables UAVs to quickly detect adversarial attacks on inertial sensors and achieve dynamic flight recovery. MARS features an attack-resilient state estimator based on the Extended Kalman Filter, which incorporates position, velocity, heading, and rotor speed measurements to reconstruct accurate attitude and angular velocity information for UAV control. Moreover, a statistical anomaly detection system monitors IMU sensor data, raising a system-level alert if an attack is detected. Upon receiving the alert, a multi-stage dynamic flight recovery strategy suspends the ongoing mission, stabilizes the drone in a hovering condition, and then resumes tasks under the resilient control. Experimental results in PX4 software-in-the-loop environments as well as real-world MARS-PX4 autopilot-equipped drones demonstrate the superiority of our approach over existing IMU-defense frameworks, showcasing the ability of the UAVs to survive attacks and complete the missions.

Figures (21)

• Figure 1: UAV onboard sensing and control architecture highlighting physical vulnerabilities of the IMUs: A) Acoustic resonant; B) Electromagnetic interference attack.
• Figure 2: MARS architecture with resilient sensor hubs (i.e., all sensors except the IMU), AD, and multi-stage flight recovery, operating at two different modes: performance-focused standard control and security-focused resilient control mode.
• Figure 3: Model-based resilient state estimator architecture.
• Figure 4: Multi-stage dynamical recovery for arbitrary autonomous missions under the coordination of high-level command, AD status and control mode selection.
• Figure 5: MARS-PX4 autopilot architecture. Gray blocks: PX4 default modules; blue blocks: our customized or specially tailored PX4 modules and communication bridges for MARS compatibility in the PX4 autopilot ecosystem.