Vulnerability in DJI drones may reveal pilot’s location

Serious security vulnerabilities have been identified in multiple DJI drones. These weaknesses had the potential to allow users to modify crucial drone identification details such as its serial number and even bypass security mechanisms that enable authorities to track both the drone and its pilot. In special attack scenarios, the drones could even be brought down remotely in flight.

DJI drones location

Photo by: RUB, Marquard

The team headed by Nico Schiller of the Horst Görtz Institute for IT Security at Ruhr University Bochum, Germany, and Professor Thorsten Holz, from the CISPA Helmholtz Center for Information Security, have presented their findings at the Network and Distributed System Security Symposium.

The researchers informed DJI of the 16 detected vulnerabilities prior to releasing the information to the public. In the course of the responsible disclosure process the manufacturer has fixed these issues.

DJI drones put to the test

The team tested three DJI drones of different categories: the small DJI Mini 2, the medium-sized Air 2, and the large Mavic 2. Later, the IT experts reproduced the results for the newer Mavic 3 model as well.

They fed the drones’ hardware and firmware a large number of random inputs and checked which ones caused the drones to crash or made unwanted changes to the drone data such as the serial number – a method known as fuzzing. To this end, they first had to develop a new algorithm.

“We often have the entire firmware of a device available for the purpose of fuzzing. Here, however, this was not the case,” as Nico Schiller describes this particular challenge. Because DJI drones are relatively complex devices, the fuzzing had to be performed in the live system.

“After connecting the drone to a laptop, we first looked at how we could communicate with it and which interfaces were available to us for this purpose,” says the researcher from Bochum. It turned out that most of the communication is done via the same protocol, called DUML, which sends commands to the drone in packets.

Four severe errors

The fuzzer developed by the research group thus generated DUML data packets, sent them to the drone and evaluated which inputs caused the drone’s software to crash. Such a crash indicates an error in the programming.

“However, not all security gaps resulted in a crash,” says Thorsten Holz. “Some errors led to changes in data such as the serial number,” Holz added.

To detect such logical vulnerabilities, the team paired the drone with a mobile phone running the DJI app. They could thus periodically check the app to see if fuzzing was changing the state of the drone.

All four tested DJI drone models were found to have security vulnerabilities. In total, the researchers documented 16 vulnerabilities. The DJI Mini 2, Mavic Air 2, and Mavic 3 models had four serious flaws.

For one, these bugs allowed an attacker to gain extended access rights in the system.

“An attacker can thus change log data or the serial number and disguise their identity,” explains Thorsten Holz. “Plus, while DJI does take precautions to prevent drones from flying over airports or other restricted areas such as prisons, these mechanisms could also be overridden,” Holz continued.

Furthermore, the group was able to crash the flying drones mid-air.

In future studies, the Bochum-Saarbrücken team intends to test the security of other drone models as well.

Transmitting unencrypted location data

In addition, the researchers examined the protocol used by DJI drones to transmit the location of the drone and its pilot so that authorised bodies – such as security authorities or operators of critical infrastructure – can access it.

By reverse engineering DJI’s firmware and the radio signals emitted by the drones, the research team was able to document the tracking protocol called DroneID for the first time.

“We showed that the transmitted data is not encrypted, and that practically anyone can read the location of the pilot and the drone with relatively simple methods,” concludes Nico Schiller.

Don't miss