Robust RF Fingerprint Identification Scheme based on Multi-Feature Fusion

With the exponential growth of wireless access devices, the issue of secure access to wireless networks has become increasingly urgent. Radio frequency (RF) fingerprint identification (RFFI), based on device hardware impairments, is considered a reliable and promising solution for device identification. However, in practical applications, the reliability of RFFI technology is inevitably affected by noise. To address these challenges, we propose a robust RFFI scheme based on multi-domain fusion feature. Specifically, we introduce a multi-domain feature fusion framework that emphasizes carrier frequency offset (CFO) as a key feature, complemented by features from other two domains. For the proposed fusion framework, we design a multi-domain fusion fingerprint extractor based on neural network. To evaluate the effectiveness of our proposed scheme in practical applications, experiment evaluation is performed based on 30 ZigBee nodes and a USRP B210 device. The results indicate that compared to existing RFFI schemes, the proposed multi-domain fusion feature scheme achieves higher identification accuracy under medium to low signal-to-noise ratio (SNR) conditions and is more robust to noise influence.