A Collaborative Lightweight Malicious Drone Detection System in Sensitive Areas

Unmanned Aerial Vehicles (UAVs) have become a critical necessity in safety and non-safety applications, including aerial photography, surveying, and delivery services. However, increased usage of UAV has raised concerns about potential malicious activities such as drones carrying drugs, carrying bombs, drones crossing the international border, etc. Detecting and mitigating the risks posed by malicious drones has become vital for ensuring safety and security in sensitive areas, especially in public places. In this paper, a two-stage approach is proposed for the detection of malicious drones. In the first stage, the machine learning algorithm is used to differentiate the drones from the other flying objects. Further, it classifies the drones with or without payload. In the second stage, using a novel heuristic method, the malicious drones are differentiated from the normal drones. The experiments were tested with both the benchmark dataset and the real-time datasets. From the experiments that were conducted in real-time using Tello drones and DJI Mavic Pro drones, it is evident that the proposed two-stage approach demonstrates high accuracy in identifying malicious drones instantly. The algorithm achieved 98.9% accuracy for drones without payloads, 93.23% for drones with payloads, and 100% accuracy in the second stage. The system’s reliability and versatility allow its application in border control and military base security.