Radio frequency fingerprint identification method based on syntactic recognition
To address the difficulty in feature extraction due to channel characteristics and noise influence, a method based on syntactic pattern recognition was proposed. Using the 802.11 Wi-Fi signal as the object, a hierarchical syntactic model was constructed based on the folding features of the radio fre...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
POSTS&TELECOM PRESS Co., LTD
2025-04-01
|
| Series: | 网络与信息安全学报 |
| Subjects: | |
| Online Access: | http://www.cjnis.com.cn/thesisDetails#10.11959/j.issn.2096-109x.2025024 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | To address the difficulty in feature extraction due to channel characteristics and noise influence, a method based on syntactic pattern recognition was proposed. Using the 802.11 Wi-Fi signal as the object, a hierarchical syntactic model was constructed based on the folding features of the radio frequency fingerprint (RFF) amplitude spectrum to achieve feature compression. The local form of the amplitude spectrum was summarized into four elements: peak, valley, upper flat shoulder, and lower flat shoulder. Three layers of syntactic mapping rules were established. The first layer extracted single line segment features (straight, rise, fall), the second layer combined adjacent segments to generate double line segment features (e.g., peak and valley), and the third layer integrated characteristic position and level value. Additionally, a subgraph fusion algorithm was proposed. Through embedding, pruning, and topology rearrangement, various syntactic models of similar equipment were integrated. Ultimately, the RFF features were compressed by 60% to 66% (from 50 dimensions to 30-33 dimensions). Experiments using 10 ESP32-WROOM 32U modules demonstrate that with a signal-to-noise ratio of 25 dB, the accuracy of the K-nearest neighbor and naive Bayes classifiers is improved by 4.3% and 5.0%, respectively. Both the intra-class error rate and inter-class error rate of open set identification are below 10.72%. This method can reduce data dimensionality while maintaining high discriminability, offering a novel approach for RFF identification. |
|---|---|
| ISSN: | 2096-109X |