GPS Spoofing Attacks on Autonomous Navigation Systems for UAVs

초록

Recently, autonomous navigation systems in UAV have been rapidly adopted. however, there is a lack of research on GPS spoofing attack against autonomous navigation for UAVs. Research on GPS spoofing attacks against UAVs has mainly focused on causing incorrect location displacement. However, this paper proposes a threat model at planning module layer for GPS spoofing attacks. This paper analyzes vulnerable structural characteristics that may arise in autonomous navigation architectures for UAVs. For this, we examine the potential for attacks in recent DRL-based autonomous navigation by exploiting these vulnerabilities. This paper first develop an attack scenario and attacker modeling for DRL-based navigation system via GPS spoofing attack. Next, using a basic common architecture of DRL-based systems, we experimentally demonstrate that DRL-based navigation system can be disrupted when inputs are spoofed. Finally, since there are constraints by EKF sensor fusion detection on altering UAV's internal state through GPS spoofing attack, this paper experimentally validates that DRL-based system can be disturbed under constrained input conditions. In particular, in experiments with constrained inputs, the feasibility of attacks was demonstrated. This paper discussed the feasibility of attacks through attack modeling and experiments of GPS spoofing attacks at autonomous navigation for UAVs. Additionally, this paper proposed methodologies for countermeasures. This work can broaden future research on GPS spoofing attacks against UAVs, particularly in modeling attacks on autonomous navigation systems in UAVs.

키워드