RF Signal-Based Multipurpose UAV Surveillance System Using Deep Neural Network

Drones, also known as Unmanned Aerial Vehicles (UAV), have recently become increasingly popular due to substantial cost reductions and technological advancements. Commercial drones are used in malicious activities that aim to harm people and societies, raising serious privacy, safety, and security concerns despite the fact that they have demonstrated their effectiveness in many common applications like filmmaking, agricultural production monitoring, and search and rescue operations. In this paper, we study about 17 types of drone classification problems using radio frequency signals transmitted during a live communication session between the drone and its controller. The study's findings demonstrated the efficacy of MobileNet V2, a Convolution Neural Network (CNN) for drone identification, with accuracy and an F1 score of above 94%.