RFUAV: A Benchmark Dataset for Unmanned Aerial Vehicle Detection and Identification

TL;DR

RFUAV addresses the lack of a comprehensive RF-based UAV identification benchmark by providing a large, real-world dataset (~1.3 TB) from 37 UAVs collected with USRPs and offering a tunable SNR framework and open evaluation tools. The authors define an RF drone fingerprint from frequency-hopping and image-bandwidth characteristics and propose a two-stage detection-identification pipeline that combines spectrogram preprocessing with contemporary deep models. They report that their preprocessing and evaluation tools achieve state-of-the-art performance on RFUAV data and show the importance of CMAP choice and frequency resolution under varying SNRs. The dataset and baseline methods are publicly available, enabling standardized comparisons and future extensions to cover more drone models.

Abstract

In this paper, we propose RFUAV as a new benchmark dataset for radio-frequency based (RF-based) unmanned aerial vehicle (UAV) identification and address the following challenges: Firstly, many existing datasets feature a restricted variety of drone types and insufficient volumes of raw data, which fail to meet the demands of practical applications. Secondly, existing datasets often lack raw data covering a broad range of signal-to-noise ratios (SNR), or do not provide tools for transforming raw data to different SNR levels. This limitation undermines the validity of model training and evaluation. Lastly, many existing datasets do not offer open-access evaluation tools, leading to a lack of unified evaluation standards in current research within this field. RFUAV comprises approximately 1.3 TB of raw frequency data collected from 37 distinct UAVs using the Universal Software Radio Peripheral (USRP) device in real-world environments. Through in-depth analysis of the RF data in RFUAV, we define a drone feature sequence called RF drone fingerprint, which aids in distinguishing drone signals. In addition to the dataset, RFUAV provides a baseline preprocessing method and model evaluation tools. Rigorous experiments demonstrate that these preprocessing methods achieve state-of-the-art (SOTA) performance using the provided evaluation tools. The RFUAV dataset and baseline implementation are publicly available at https://github.com/kitoweeknd/RFUAV/.

Figures (13)

• Figure 1: Schematic representation of the RF-based UAV monitoring framework. The system captures and analyzes radio frequency signals emitted by UAV operators to detect and classify drone activity. The monitoring station processes the received signals, facilitating UAV detection and identification of drone types.
• Figure 2: Hardware architecture and interconnections of the RF signal acquisition platform. The system integrates multiple components, including a GNU Radio controller, USRP devices, a spectrum analyzer, and an AU monitoring unit. These components facilitate high-quality RF signal collection, processing, and analysis.
• Figure 3: Statistical overview of the original dataset. The figure presents key attributes of the collected RF data, including file size (GB), signal-to-noise ratio (SNR) in dB, and center frequency (MHz × $10^{-1}$) for various UAV controllers.
• Figure 4: Comprehensive statistical analysis of key features observed in drone communication signals. Left Polar Plot: Displays the frequency-hopping signal bandwidth (FHSBW, MHz), image transmission signal bandwidth (VTSBW, MHz), and frequency-hopping signal duration time (FHSDT, ms). Right Polar Plot: Depicts the frequency-hopping duty cycle (FHSDC, ms) and frequency-hopping signal periodicity (FHSPP, ms).
• Figure 5: Main distribution of image transmission signal duration in the RFUAV