Authentication by Intelligent Learning: A Novel Hybrid Deep Learning/Machine-Learning Radio Frequency Fingerprinting Scheme

Abstract

LoRa technology, widely used in the Internet of Things (IoT) domain, faces challenges with traditional cryptographic authentication methods due to power constraints and computing overhead. Radio Frequency Fingerprinting (RFFI) emerges as a low-cost, low-power solution. In this paper, we propose a novel RFFI method for authenticating LoRa devices, which combines deep learning (DL) features and supervised machine learning (ML) for classification. We examine authentication performance across multiple LoRa Spreading Factors (SFs) and assess the hybrid DL/ML approach. Moreover, we introduce a novel IQ sample transformation method by utilizing the histogram of the IQ samples and the 3D image channels. Among the DL models explored, the SwinTransformer (ST) and Few Shots Learning (FSL) models stand out. Experimental results show that our system achieves 97.5% accuracy with reduced complexity compared to the baseline schemes.