RF-NeuralNet: Lightweight Deep Learning Framework for Detecting Rogue Drones from Radio Frequency Signatures

Unmanned aerial vehicles (UAVs) have emerged as a revolutionary technology with diverse applications in fields such as crop monitoring, logistics, and traffic surveillance. Despite all these advantages, they also pose certain challenges such as privacy breaches, potential collision risks with airplanes, and terrorism activities. To mitigate these concerns, various techniques have been developed for UAV detection. In this paper, we propose a computationally efficient deep learning network RF-NeuralNet for UAV detection and mode identification using RF fingerprints. The proposed network involves a multiple-level skip connection to mitigate the gradient vanishing problem and multiple-level pooling layers for deep-level feature extraction. We evaluate the performance of the proposed RF-NeuralNet based on multiple UAV monitoring tasks (i.e., UAV identification, classification, and operational mode). Our proposed framework outperformed other state-of-the-art models and achieved an overall accuracy of 89%.