UC Irvine · Irvine, CA
Security analysis of camera- and AI-based drone autonomy, from target tracking and precision landing to vision-language models.
UC Irvine’s Autonomous & Smart Systems Guard (AS²Guard) research group, led by Alfred Chen, studies the security of AI in the latest small-UAS autonomy functions. Its FlyTrap study of autonomous target tracking was presented at NDSS 2026, responsibly disclosed to the affected drone makers, and covered by more than 15 news outlets; its MirageMarker study of marker-based autonomous landing appears at IROS 2026. With GWU and TSU, the group is developing defenses for vision-language-model-enabled drones and a drone-delivery benchmark.
Research thrusts
The first security analysis of camera-based autonomous target tracking in drones, demonstrated against commercial systems and presented at NDSS 2026.
The first security study of fiducial-marker pose estimation for precision landing, evaluated against the PX4 autopilot’s native landing mode in NVIDIA Isaac Sim (IROS 2026).
Ongoing work on defending vision-language models used for drone tracking and landing against visual prompt injection.
Figures
Demos
UC Irvine’s demonstration of FlyTrap, the first security analysis of camera-based autonomous target tracking in drones. A specially patterned umbrella draws a tracking drone dangerously close. Evaluated on commercial drones and presented at NDSS 2026.
A real-world capture of marker-based pose estimation, comparing the estimate from a clean marker with one under the MirageMarker perturbation.
Subtle geometric changes to a fiducial landing-marker board shift a PX4 drone’s precision landing 8–12 m off the pad in NVIDIA Isaac Sim, while clean boards land within centimeters. Presented at IROS 2026.
Use cases
People
Publications
2 published or accepted papers
Network and Distributed System Security Symposium (NDSS), 2026
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2026