Human Augmentation of UAV Cyber-Attack Detection

Haibei Zhu, Mahmoud Elfar, Miroslav Pajic, Ziyao Wang, Mary L. Cummings

Abstract: Unmanned aerial vehicles (UAVs) have extensive applications in both civilian and military applications. Nevertheless, the continued development of UAVs has been accompanied by security concerns. UAV navigation systems are potentially vulnerable to malicious attacks that target their Global Positioning System (GPS). Thus, efficient GPS hacking detection with high success rate is paramount. Significant effort has been put into developing autonomous hacking detection techniques. However, little research has considered how a human operator can contribute to the security of such systems. In this paper, we propose a human-autonomy collaborative approach for a single operator of multiple-UAV supervisory control systems, where human geo-location is used to help detect possible UAV cyber-attacks. An experiment was designed and conducted using the RESCHU-SA experiment platform to evaluate this approach. The primary results show that 65% of all experiment sessions reached over 80% success rate in UAV hacking detection, while only 17% of participants lost one or more UAVs because of incorrect hacking detections. These results suggest that such an approach could help achieve better security guarantees for human-in-the-loop autonomous UAV systems that are prone to cyber-attacks.