Researchers at the University of California Irvine (UC Irvine) have discovered a vulnerability in autonomous drones that target and track their targets using Artificial Intelligence through cameras.
With a specially patterned umbrella, drone tracking systems can be fooled by dragging them along the ground, allowing the target to catch or throw them to the ground.
Autonomous consumer drones, as well as professional drones used for security surveillance, border control, or law enforcement, often use Artificial Intelligence systems through cameras.
These systems allow drones to track people or other objects such as vehicles. However, tracking targets based solely on visual recognition poses significant security problems, as the researchers demonstrate in the study “ FlyTrap: Physical Distance-Pulling Attack Towards Camera-based Autonomous Target Tracking Systems ,” which was published as a draft on Arxiv. The research was presented this week at the Network and Distributed System Security Symposium in San Diego.
Scientists at UC Irvine found that these drones are vulnerable to attacks that increase distance. The researchers used umbrellas with specially designed optical patterns to fool artificial intelligence tracking systems.
They call the system FlyTrap.
The drone's tracking system believes that the person holding the patterned umbrella is moving away, even though it remains in the same place. According to the erroneous data received by the drone, the person continues to approach and can thus be caught in a net or thrown to the ground.
The researchers show that FlyTrap can be used to attract and then disable police drones or drones on security flights. The scientists tested FlyTrap with three commercially available drones equipped with tracking systems: the DJI Mini 4 Pro and DJI Neo, and the Hover Air X1.
All three drones were vulnerable to the attack. The scientists demonstrated that they could affect a drone under different weather and lighting conditions without requiring special electronic equipment.
DOI: https://doi.org/10.48550/arXiv.2509.20362
Although the press releases will range from very select to rare, I said I'd pass...because sometimes the editors hide.
