Shaoyuan Xie/UC Irvine
Researchers have discovered that there are certain patterns on umbrellas that can confuse systems and attract drones, allowing them to be destroyed.
Researchers in the United States have demonstrated an unusual but surprisingly simple method for tricking autonomous target-tracking drones: an umbrella with a print. The technique exploits vulnerabilities in artificial intelligence systems that allow drones to automatically follow people or objects.
The study, led by a team at the University of California, Irvine, presents a method called FlyTrap. According to the researchers, the method can trick the dronescausing them to approach a target until they are captured with a net or forced to fall.
Autonomous tracking technology, often marketed as “active tracking” or “dynamic tracking,” relies on cameras and neural network algorithms to identify and track moving objects without direct human control.
However, the UC Irvine team found that these screening systems can be fooled by specially designed visual patternss. In his experience, a person simply opens an umbrella covered with a unique graphic design. The drone’s tracking algorithm mistakenly interprets the pattern as an object moving away from the aircraft.
To maintain what it believes is a consistent following distance, the drone automatically flies toward the umbrella. By repeatedly adjusting your position, you end up close enough for the operator to capture it with a net or cause a controlled collision, says .
“Autonomous target tracking represents enormous potential and significant risk,” said Alfred Chen, assistant professor of computer science at UC Irvine and co-author of the study. While security agencies and law enforcement organizations are adopting the technology, criminals have also been using tracking drones for stalking and other malicious activities.
The researchers tested their method on three commercially available drones: the DJI Mini 4 Pro, the DJI Neo and the HoverAir X1. In all cases, drones could be lured to close enough to be captured or forced to fall. The team claimed to have disclosed the vulnerabilities to the manufacturers involved.
The findings raise concerns about the reliability of autonomous drone systems in security-sensitive environments. For example, the method could allow criminals disable surveillance drones or bypass monitoring near borders or restricted areas.
At the same time, the researchers emphasize that the technique can also serve as a defense tool. People who are being harassed or stalked by drones could potentially use similar tactics to disable the devices.
