16 October 2024 in IET Communications
Radio frequency fingerprint identification (RFFI) is a widely used technique for authenticating equipment. It identifies transmitters by extracting hardware defects found in the RF front end. Recent research has focused on the impact of transmitters and wireless channels on radio frequency fingerprint (RFF). Most work is based on the same receiver assumption, while the influence of the receiver on RFF remains unresolved. This paper focuses on the impact of receiver hardware characteristics on RFF and proposes a few‐shot cross‐receiver RFFI method based on feature separation. Data augmentation with noise addition and simulated channels addresses sparse sample issues and enhances the model's robustness to channel variations. Simultaneously, feature separation is realized by reducing the correlation between transmitter and receiver features through classification loss and similarity loss. We evaluate the proposed approaches using a large‐scale WiFi dataset. It is shown that when a trained transmitter classifier is deployed on new receivers with only 30 samples per trained transmitter, the average identification accuracy of the proposed method is 83.6%. This accuracy is 9.45% higher than the baseline method without considering transmitter hardware influence. After fine‐tuning, the average identification accuracy can reach 98.25%.
13 May 2025 in IEEE Communications Letters
Domain adaptation has been widely applied in radio frequency fingerprint identification (RFFI) to address the domain shift problem. However, conventional methods rely on both source and target domain data, presenting practical challenges related to data privacy and big data transmission. This study proposes a source-free domain adaptation method enabling target model adaptation without source data. Specifically, our approach utilizes multi-neighborhood semantic consistency and spatial adjacency to encourage uniform predictions for features with analogous characteristics, serving as regularization objectives during optimization. Additionally, a discrete loss function distracts the prediction of potentially dissimilar features. Experimental results show that this method achieves an average identification rate of over 89% without relying on source domain data.