Analysis of interactions with remotely controlled devices has been used to detect the onset of hijacking attacks, as well as for forensics analysis, e.g., to identify the human controller. Its effectiveness is known to depend on the remote device type as well as on the properties of the remote control signal. This paper shows that the radio control signal sent to an unmanned aerial vehicle (UAV) using a typical transmitter can be captured and analyzed to identify the controlling pilot using machine learning techniques. Twenty trained pilots have been asked to fly a high-end research drone through three different trajectories. Control data have been collected and used to train multiple classifiers. Best performance has been achieved by a random forest classifier that achieved accuracy around 90% using simple time-domain features. Extensive tests have shown that the classification accuracy depends on the flight trajectory and that the pitch, roll, yaw, and thrust control signals show different levels of significance for pilot identification. This result paves the way to a number of security and forensics applications, including continuous identification of UAV pilots to mitigate the risk of hijacking.
Drone Pilot Identification by Classifying Radio-Control Signals / A. Shoufan, H.M. Al-Angari, M.F.A. Sheikh, E. Damiani. - In: IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY. - ISSN 1556-6013. - 13:10(2018 Oct), pp. 2439-2447.
Drone Pilot Identification by Classifying Radio-Control Signals
E. DamianiUltimo
2018
Abstract
Analysis of interactions with remotely controlled devices has been used to detect the onset of hijacking attacks, as well as for forensics analysis, e.g., to identify the human controller. Its effectiveness is known to depend on the remote device type as well as on the properties of the remote control signal. This paper shows that the radio control signal sent to an unmanned aerial vehicle (UAV) using a typical transmitter can be captured and analyzed to identify the controlling pilot using machine learning techniques. Twenty trained pilots have been asked to fly a high-end research drone through three different trajectories. Control data have been collected and used to train multiple classifiers. Best performance has been achieved by a random forest classifier that achieved accuracy around 90% using simple time-domain features. Extensive tests have shown that the classification accuracy depends on the flight trajectory and that the pitch, roll, yaw, and thrust control signals show different levels of significance for pilot identification. This result paves the way to a number of security and forensics applications, including continuous identification of UAV pilots to mitigate the risk of hijacking.File | Dimensione | Formato | |
---|---|---|---|
drone.pdf
accesso riservato
Tipologia:
Publisher's version/PDF
Dimensione
1.61 MB
Formato
Adobe PDF
|
1.61 MB | Adobe PDF | Visualizza/Apri Richiedi una copia |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.