Emphasized Spectrum-Based BLE Signal Fingerprinting Dataset for IoT Device Authentication

This dataset contains emphasized spectral feature data extracted from Bluetooth Low Energy (BLE) signals for IoT device authentication research.
The data were generated as part of the experimental evaluation presented in the accompanying paper, which proposes a signal fingerprinting framework based on emphasized frequency regions to capture hardware-dependent characteristics of BLE transmission devices.
The dataset is organized by filter bank type applied to the emphasized frequency regions. Each top-level directory corresponds to a specific filter bank configuration (e.g., Linear Scale, Mel Scale, Inverse Mel Scale, and Gammatone Scale). Within each directory, subfolders contain data associated with individual BLE transmission modules.
Each NumPy (.npy) file represents emphasized spectral features extracted from a single BLE module. For each module, multiple BLE signals were collected and processed through framing, FFT, and emphasized filter bank application. The resulting feature matrices are stored by vertically stacking the features from multiple signals into a single NumPy array.
Specifically, each .npy file contains emphasized spectral features derived from 500 BLE signals transmitted by the same module. By partitioning the stored array into equal segments, users can recover the emphasized spectral representation corresponding to each individual signal.
This dataset is intended to support reproducibility and comparative evaluation of BLE signal fingerprinting, anomaly detection, and device authentication methods at the physical layer. While the source code used to generate the dataset is not publicly released due to security and intellectual property considerations, the dataset provides sufficient structure and documentation for independent analysis and validation.macdd l