Autonomous Systems
Navigation
Autonomous drones use computer vision, SLAM, and AI for navigation. These systems can be attacked through sensor spoofing and adversarial inputs.
Autonomous Navigation Systems
Visual Navigation
- • Optical flow (position hold)
- • SLAM (mapping)
- • Obstacle avoidance
- • Landing detection
Attack Vectors
- • Camera blinding
- • Pattern spoofing
- • Adversarial patches
- • Sensor flooding
Vision System Attacks
vision-attacks.sh
# Attacks on visual navigation:
# 1. IR blinding (DJI sensors)
# IR LEDs can saturate downward sensors
# Causes altitude hold failure
# 2. Optical flow confusion
# Moving patterns under drone
# Causes position drift
# 3. Obstacle detection spoofing
# Fake obstacles to block path
# Or mask real obstacles
# 4. Landing marker spoofing
# Fake AprilTags or landing markers
# Redirect precision landing
# 5. SLAM manipulation
# Modify visual features
# Confuse loop closure detection
# ROS-based drone simulation
roslaunch px4 mavros_posix_sitl.launch
# Test attacks in simulation first