Design and development of an SDR-based system for real-time detection and characterization of drone RF signatures

Abstract

The widespread availability of consumer drones has introduced new challenges related to safety, security, and privacy, as these platforms are increasingly misused in sensitive or restricted areas., existing counter-drone technologies–such as radar, optical tracking, and multi-sensor fusion–offer reliable performance but are often prohibitively expensive and impractical for large-scale civilian deployment. This work presents a low-cost framework for real-time drone detection and classification that leverages the radio frequency (RF) emissions exchanged between drones and their controllers. The system is built on a software-defined radio (SDR) platform (USRP B210), which captures RF signals and converts them into spectrograms for analysis using deep learning. A labeled dataset of drone and non-drone signals was developed to train and evaluate detection models. Two state-of-the-art architectures, YOLOv5 and Faster R-CNN, were adapted to this task, with evaluation under varying signal-to-noise ratio (SNR) conditions. Results demonstrate that the proposed system achieves high detection accuracy and robustness even in noisy environments, highlighting its potential as a scalable and practical solution for RF-based drone monitoring.