Researchers at Ben Gurion University, Israel have developed a technique that can identify the location of a drone operator with almost 73% accuracy.
Unmanned Aerial Systems (UAS) commonly known as Drones posit a significant security threat enforcing organizations to guard their airspace. In the next few years, their superior features, such as agility, dexterity and widespread availability will market drones as ‘ubiquitous business tools’ risking exposure and compromising security barriers.
Research [PDF] conducted by Ben Gurion University of the Negev, Israel: ‘Can the operator of a drone be located by following the drone’s path’ profusely explain a growing need to adapt localization and detection methods to assuage malicious attacks and operations via drones.
The researchers extensively discuss the use of deep neural networks to localize drone operators using a realistic simulation environment that can help collect imperative data with 73% accuracy.
There are more than 1 million drone operators residing in the US alone and the global market for UAS will be worth $21.8 billion by 2027. This showcases how copiously varied institutions deploy drones not only for security purposes but for leisure too. However, this has led to an increase in malicious activities ensuing drones controlled by cybercriminals and perpetrators.
Illicit incidents revolving around drones
The Heathrow Airport incident is an insinuation of the fact, how vicariously drones were deployed for illicit activities. In April 2016, A British Airways flight from Geneva was hit by, what was believed to be by a helicopter-style drone.
A full-fledged investigation took place for security reasons as the drone did not appear on the traffic control radar as well. Another alarming incident happened in Mexico wherein, a drone transporting crystal meth crashed near a supermarket.
Similarly, an incident in Massachusetts showcased drones as a significant threat to wildlife. Moreover, a drone sighting at Frankfurt Airport, Germany, in March 2020 caused the suspension of all flights perturbing chaos and alarming security personnel.
Identifying Operators location using deep neural networks
However, the researchers of the paper came up with a rather interesting methodology ensuing the localization of operators and mitigating risk factors. They conducted an experiment using a flight simulation that realistically showcased sun gazes, obstructions, and visual effects used to identify the operator’s location given the path of the drone.
The research explains using Aerial Informatics and Robotics Simulation (AirSim) built using Epic games Unreal Engine with neural networks (sensors to recognize patterns). This helped researchers identify a minute yet critical point that is; collecting signals from unintentional maneuvers performed by pilots/operators.
However, through simulations, three viewpoints were explored including the first-person view, which helped procure the perspective of the real pilot operating the drone.
Supporting existing techniques that lack versatility
The researchers also explain how their work can complement existing radio frequency (RF) techniques. Usually locating signals posit a major challenge. There are a lot of WiFi, Bluetooth, and IoT signals present in the air. In order to curb this, the researchers explained the use of network-enabled sensors around the flight path which will help defenders locate the pilot.
The existing techniques are only limited to identifying signals from the antenna. But malicious operators can easily transfer these signals to another antenna. Their techniques using neural networks can extract direct information and that too from the operator’s viewpoint.
“Our system can now identify patterns in the drone’s route when the drone is in motion, and use it to locate the drone operator,” said lead researcher Eliyahu Mashhadi, a BGU computer science student.
The use of Neural Networks can potentially invade operators privacy
The researchers explain their technique is actually dependent on human behavior. Which includes how the operator drives, their reaction, and complex action as well. Much of their results are dependent upon the technical experience of the operator and their training level.
However, this particular technique deployed by the researchers can lead to widespread surveillance by state-owned enterprises compromising public safety. Also, analyzing human behaviors can potentially invade commercialized operators’ privacy. But existing techniques lack versatility and can only be deemed useful for specific brands only. Deep learning methods ensuing neural networks can help expand reach.
“Now that we know we can identify the drone operator location, it would be interesting to explore what additional data can be extracted from this information,” said Dr. Yossi Oren, one of the researchers. “Possible insights would include the technical experience level and even precise identity of the drone operator.”