Radio Frequency Fingerprint Identification for Security in Low-Cost IoT Devices

TL;DR

This paper introduces a transformer-based radio frequency fingerprint identification system for low-cost IoT devices, enhancing classification accuracy in low SNR conditions through data augmentation and multi-packet inference.

Contribution

It applies transformer models to RFFI, enabling variable-length signal processing and improves low SNR performance with data augmentation and multi-packet inference techniques.

Findings

Data augmentation boosts low SNR RFFI accuracy by up to 50%.

Multi-packet inference increases accuracy by over 20%.

System effectively classifies 10 LoRa devices under various SNR conditions.

Abstract

Radio frequency fingerprint identification (RFFI) can uniquely classify wireless devices by analyzing the received signal distortions caused by the intrinsic hardware impairments. The state-of-the-art deep learning techniques such as convolutional neural network (CNN) have been adopted to classify IoT devices with high accuracy. However, deep learning-based RFFI requires input data of a fixed size. In addition, many IoT devices work in low signal-to-noise ratio (SNR) scenarios but the low SNR RFFI is rarely investigated. In this paper, the state-of-the-art transformer model is used as the classifier, which can process signals of variable length. Data augmentation is adopted to improve low SNR RFFI performance. A multi-packet inference approach is further proposed to improve the classification accuracy in low SNR scenarios. We take LoRa as a case study and evaluate the system by classifying 10 commercial-off-the-shelf LoRa devices in various SNR conditions. The online augmentation can boost the low SNR RFFI performance by up to 50% and multi-packet inference can further increase it by over 20%.