Action-Driven UAV Fingerprint Verification with Perception Data

Unmanned Aerial Vehicles (UAVs) have a wide range of applications in various industries. Their navigation in challenging environments like dense forests and urban areas requires a high degree of precision. However, vulnerabilities such as external attacks, inherent bugs, and design flaws can cause erroneous execution of actions during flight, thereby increasing the potential for collisions. To ensure the accuracy and integrity of actions during flight, we propose and implement a real-time verification system, FingVer. By iteratively executing pre-defined critical actions and analyzing UAV’s perceptual data within a designated time window, FingVer generates distinct fingerprints for actions. During UAV flights, the method verifies and clusters intended actions based on the fingerprint library, ensuring accuracy and reliability. Experimental results demonstrate FingVer ’s effectiveness, achieving precision rates of 87.4% and 89.4% for parallel turning and lifting actions, respectively, with recall rates exceeding 86%.