An Israeli firm is exploring a way to track the path of a combat drone, all the way back to the operator, utilizing a deep neural network. Credit: Handout.

The growing threat of hostile combat drones, and potential drone swarms, being used for tactical advantage is causing deep concern for militaries around the world.

Russia, China and the United States all took note after a daring drone attack on oil reserves in Saudi Arabia in September of 2019 eluded missile defences and caused considerable damage to Aramco oil facilities.

Houthi forces in Yemen took responsibility for the attack, Iran denied any involvement, and the US blamed Iran. While the latter is largely considered responsible, no one is 100% sure where they were launched.

Up to now, it has been difficult to determine the source of the drones, but that might change thanks to a team of Israeli scientists who are developing a system they believe will let them accurately locate the operator of hostile drones and neutralize him, Arie Egozi of Breaking Defense reported.

Researchers at Ben Gurion University in Beer Sheva in southern Israel, Gera Weiss and Eliyahu Mashhadi, are using a realistic simulation environment to collect the path of the drone when flown from launch point and monitor its flight path, the report said.

“We insert all the points along the flight path into a deep neural network that was trained to be able to predict the exact launch point and the location of the drone operator,” Mashhadi said.

Testing the model with the flight simulator, the team were able to locate and target the drone operator 78% of the time, the report said.

Today, counter-drone systems use radio frequency to locate the operators, while using electro-optical, radar and other sensors to track the drones.

“All the approaches that we are aware of for locating operators, not just the drones, use RF sensors.”

Mashhadi explained that there are automatic and semi-automatic methods for locating the operators based on radio communications between the drone and its operator, the report said.

“There are a number of problems with this approach. Firstly, such methods are usually tailored to a specific brand of drones,” he said. “Furthermore, the radio signal can only be recorded near the drone.

“Finally, there are ways for malicious drone designers to apply cryptography and electronic warfare techniques to make localization by analysis of radio signals very difficult.”

Mashhadi explained that their experiments show the reactions of the operator due to environmental and physical conditions, give away enough information to obtain substantial information about the location of the operator by analyzing the drone’s path in the sky, the report said.

“To allow for a controlled environment, we conducted all our experiments with a flight simulator that provides a realistic flight experience for the operator that includes sun gazes, obstructions, and other visual effects that produce the reactions of the operators that allow us to identify their location,” he said.

The research team used AirSim (Aerial Informatics and Robotics Simulation), an open-source, cross-platform simulator for drones, ground vehicles and other objects, the report said.

“The neural network that we have designed was able to take advantage of these relations when we asked it to use only position or only rotation information,” Mashhadi said.

Israeli sources say a system able to find the operator in real time will become critical because, in most cases, the operator is flying more than one drone.