UCI Computer Scientists Uncover Critical Drone Vulnerability: 'FlyTrap' Attack Uses Ordinary Umbrella

University of California, Irvine (UCI) computer scientists have identified a significant security vulnerability in autonomous target-tracking drones, carrying broad implications for public safety, border security, and personal privacy. The UCI team successfully demonstrated a method for attackers to manipulate these drones using an ordinary umbrella, causing the aircraft to approach closely enough for capture or collision.

This vulnerability could allow attackers to manipulate autonomous drones using an ordinary umbrella, drawing them close enough for capture or collision.

Introducing FlyTrap: A Physical-World Attack Framework

Researchers developed a physical-world attack framework named FlyTrap, designed to exploit deficiencies in camera-based, autonomous target-tracking technology. This technology enables drones to follow targets without direct human control, utilizing AI-powered functions often called 'active track' or 'dynamic track.' These functions are widely deployed in critical applications such as border control, security surveillance, and law enforcement.

The team presented its findings and specifications for the FlyTrap attack platform at the Network and Distributed System Security Symposium in San Diego.

Expert Insight: The Dual Nature of Autonomous Tracking

Alfred Chen, UC Irvine assistant professor of computer science and paper co-author, emphasized that autonomous target tracking presents both significant potential and inherent risks. He highlighted its growing adoption by law enforcement and security agencies for border patrol and public safety, alongside its potential misuse by criminals.

"Our work represents the first comprehensive security study of this widely deployed technology," stated Alfred Chen.

Understanding the 'Distance-Pulling Attack'

Chen's research group identified a novel 'distance-pulling attack' specifically designed to draw victim drones closer to an attacker. This attack employs an ordinary umbrella covered with a specifically designed visual pattern. This pattern can effectively deceive the neural network tracking systems of autonomous drones.

The drone's computer logic misinterprets the umbrella's pattern as a target moving away, even when the umbrella is stationary. To correct what it perceives as an increasing distance, the drone approaches the umbrella holder, potentially allowing for its physical capture or causing a crash. This method distinctly aims for physical capture or collision, differing from attacks that merely result in tracking loss.

Successful Demonstrations and Disclosure

UC Irvine researchers successfully demonstrated FlyTrap attacks on several commercial drones, including the DJI Mini 4 Pro, DJI Neo, and HoverAir X1. Test results clearly indicated that an attack could draw drones close enough for capture or to cause physical crashes. The team has since reported these critical vulnerabilities to manufacturers DJI and HoverAir.

Broader Implications and Potential Uses

The research suggests diverse potential uses for these distance-pulling attacks. Criminals could exploit this technique to evade law enforcement drones or hinder unpiloted aircraft patrolling sensitive border zones. Conversely, the technique could potentially be used by individuals experiencing harassment from drones to disable them.

Call for Urgent Security Improvements

Lead author Shaoyuan Xie, a UC Irvine graduate student researcher, stressed the urgency of these findings.

"Our findings indicate an urgent need for security improvements in autonomous target-tracking systems before broader deployment in critical infrastructure," stated Shaoyuan Xie. He further suggested that the operation of such drones in public or critical security settings should be re-evaluated given the demonstrated ease of manipulation.

FlyTrap: A Local and Robust Attack

The FlyTrap attack methodology is remarkably simple, relying on the physical act of opening a portable umbrella. The system operates locally, meaning it requires no external signaling or wireless data connectivity. Furthermore, it functions robustly across various weather and lighting conditions and employs a progressive distance-pulling strategy by subtly manipulating drone-tracking algorithms.

Research Documentation

Comprehensive project documentation, including a website, datasets, metrics, demonstration videos, social media, and an extended paper, is available to support future security improvements. All drone data and experiments were concluded before December 22, 2025.