Few-Shot RF Fingerprint Authentication via Unfolded Sparse Recovery

Abstract

In this paper, we propose a few-shot authentication framework based on radio-frequency fingerprint identification (RFFI). Unlike conventional RFFI approaches, which often memorize device-specific features, our framework learns to accurately compare over-the-air (OTA) signals from different Internet of Things (IoT) devices, enabling effective authentication with limited training samples. The proposed architecture integrates a transformer encoder with deep-unfolded alternating direction method of multipliers (ADMM) layers. The transformer encoder maps raw in-phase/quadrature (IQ) samples extracted from OTA signals into channel-invariant fingerprint vectors, while the ADMM layers further refine these vectors through learned group-sparse shrinkage operations. The entire framework is trained end-to-end using episodic training, allowing it to authenticate previously unseen devices using only a few reference signals without retraining. We evaluate the proposed approach on a public LoRa dataset comprising 60 devices and a custom dataset collected from 10 commercial Waveshare SX126X devices using a USRP B210 receiver. Experimental results on both datasets demonstrate the effectiveness of the proposed method.