A UAV Testbed for Diagnosing Hardware Vulnerabilities: Quantifying Sim-to-Real Discrepancies in PX4 Flight Logs

Area of Science:

• Robotics and Control Systems
• Cyber-Physical Systems Security
• Aerospace Engineering

Background:

• Autonomous Unmanned Aerial Vehicle (UAV) systems require robust security validation.
• Existing simulation environments often fail to capture real-world hardware complexities and environmental factors.
• Cyber-physical vulnerabilities in UAVs pose significant risks to safety-critical operations.

Purpose of the Study:

• To establish a comprehensive UAV testbed for quantitative hardware vulnerability diagnosis.
• To validate the cyber-physical security of autonomous UAVs by comparing simulation and real-world flight data.
• To create a foundational framework for anomaly detection and security validation in UAV systems.

Main Methods:

• Leveraging comparative flight logs from Software-In-The-Loop (SITL) simulations and real-world quadrotor missions.
• Utilizing a unified data pipeline with the uORB message bus and ULog format for high-resolution telemetry extraction (IMU, state-estimation, actuator control).
• Performing side-by-side time-series and statistical analyses of simulated versus real-world data across varying environmental conditions.

Main Results:

• Identified critical sim-to-real discrepancies in sensor fidelity, GPS interference, and onboard resource behavior.
• Quantified hardware-induced noise, mechanical vibrations, and electromagnetic disturbances impacting flight stability and reliability.
• Established mathematical methods (variance, probability distribution shifts) to distinguish physical variability from anomalous or adversarial behavior.

Conclusions:

• The developed UAV testbed provides a rigorous baseline for assessing cyber-physical security.
• The findings highlight the necessity of real-world data for accurate vulnerability diagnosis and security validation.
• This framework supports the development of robust anomaly detection models for secure autonomous UAV operations.