FlyTrap Attack on Autonomous Drones (ics.uci.edu)

Researchers at the University of California, Irvine have unveiled a significant security vulnerability in autonomous target-tracking drones, demonstrating an attack method they call FlyTrap. This novel technique allows attackers to manipulate the drones using a simple umbrella adorned with a specific visual pattern. By deceiving the drones' neural networks into thinking the umbrella holder is moving farther away, the drones are drawn closer, making them vulnerable to capture or crashes. This research highlights risks associated with the increasing reliance on AI-driven tracking systems in security and law enforcement applications, raising concerns over both public safety and privacy. The FlyTrap attack underscores the urgent need for enhanced security measures in autonomous drone technology, particularly as these systems are deployed in sensitive areas like border control and surveillance. The team's tests successfully demonstrated this vulnerability on three commercial drones, revealing how easily drones can be compromised in real-world scenarios without the need for external signals. Co-author Alfred Chen emphasized that while these technologies promise great potential, the findings call for a reconsideration of their deployment in critical infrastructure due to the risks involved. The researchers have responsibly disclosed their findings to drone manufacturers in hopes of enhancing system safety before wider use.