RF-Fingerprint-BT-IoT: Real-world Frequency Hopping Bluetooth dataset from IoT devices for RF fingerprinting

A real-world radio frequency (RF) fingerprinting dataset for commercial off-the-shelf (COTS) Bluetooth emitters under challenging testbed setups is presented in this dataset. The dataset includes emissions from 10 COTS IoT emitters (2 laptops and 8 commercial chips) that are captured with a National Instruments Ettus USRP X300 radio outfitted with a UBX160 daughterboard and a VERT2450 antenna. The receiver is tuned to record a 2 MHz bandwidth of the spectrum centered at the 2.414 GHz frequency. This is a first-of-its-kind dataset for fingerprinting Bluetooth emitters under challenging and diverse indoor laboratory setups. The dataset is split into two:  Day1 and Day2 each of which is recorded in a different time frame, location, and testbed setup to enable critical generalization test of the trained deep learning (DL) model. The authors suggest training the DL model with the Day1 dataset which is a simpler setup with recordings under varied Bluetooth signal strengths followed by evaluating the generalization capability of the model with Day2 dataset which is a challenging and vastly different setup compared to Day1. This evaluation will validate the real-world deployment capability of the trained DL model. The dataset follows the SigMF specifications with certain field extensions to facilitate the fingerprinting application and include additional metadata fields. Each capture is of length 40 Mega Samples (MS) and is associated with a JSON metadata file.   

本数据集提供了一款面向商用现货（Commercial Off-The-Shelf, COTS）蓝牙发射设备的真实射频（Radio Frequency, RF）指纹数据集，其测试床搭建环境极具挑战性。该数据集包含10台COTS物联网（Internet of Things, IoT）发射设备的射频发射信号，其中包括2台笔记本电脑与8款商用芯片。信号采集采用美国国家仪器（National Instruments）旗下Ettus USRP X300无线电平台，搭配UBX160子板与VERT2450天线完成捕获。接收机调谐至以2.414 GHz为中心频率的频段，采集带宽为2 MHz的频谱数据。本数据集是首款针对挑战性与多样化室内实验室环境的蓝牙发射设备射频指纹数据集，属同类研究首例。数据集分为Day1与Day2两个子集，二者分别在不同的时间周期、采集地点与测试床配置下录制，用于验证训练得到的深度学习（Deep Learning, DL）模型的泛化性能。作者建议使用Day1数据集训练深度学习模型——该数据集的测试环境更为简单，且覆盖了不同强度的蓝牙信号——随后使用Day2数据集评估模型的泛化能力。与Day1相比，Day2的测试环境更具挑战性且差异显著，该评估流程可验证训练完成的深度学习模型的真实部署能力。本数据集遵循SigMF规范，并针对射频指纹应用扩展了部分字段，同时包含额外的元数据字段。每段采集数据的长度为40百万采样点（Mega Samples, MS），并配套对应的JSON元数据文件。