Threat Modelling UxV's

Autonomous and Uncrewed Systems (UxVs)—encompassing UAVs, UGVs, and UUVs—face a unique and rapidly evolving landscape of cyber-physical threats. This project outlines a comprehensive threat modelling framework specifically tailored to the architectures of modern autonomous systems.

Identified Threat Vectors

• Sensor Spoofing & Blinding: Attacks targeting LiDAR, radar, and optical sensors to feed false environmental data to the autonomous navigation stack.
• GNSS Spoofing & Jamming: Disrupting or manipulating Global Navigation Satellite System signals to alter the vehicle’s perceived location, potentially leading to hijacking or catastrophic failure.
• Adversarial Machine Learning: Utilizing adversarial patches or specifically crafted inputs to deceive the computer vision models responsible for object detection and classification.
• Middleware Exploitation (ROS/ROS2): Exploiting misconfigured Data Distribution Service (DDS) instances or unauthenticated ROS nodes to inject malicious commands or eavesdrop on internal communications.
• Supply Chain & Hardware Implants: Risks associated with compromised hardware components and malicious firmware updates.

Defense Strategies

The framework details mitigation strategies including robust sensor fusion, zero-trust architectures for robotic middleware, cryptographic signing of firmware, and the implementation of adversarial training for perception models. By systematically addressing these vectors, the resilience of autonomous systems can be significantly enhanced.