A radio frequency fingerprinting method for wireless node identity authentication
By converting wireless signals into time-spectrum graphs and training them using feature projection and alignment-free feature distillation loss from teacher and student identification models, the feature extraction challenge for wireless node authentication in short-frame wireless signal scenarios is solved, achieving efficient authentication and lightweight deployment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHANGCHUN UNIV OF SCI & TECH
- Filing Date
- 2026-05-26
- Publication Date
- 2026-06-23
AI Technical Summary
Existing radio frequency fingerprinting methods struggle to effectively extract the identity features of wireless nodes in short-frame wireless signal scenarios, and traditional distillation methods fail to establish effective supervisory relationships between heterogeneous models, resulting in difficulties in deploying recognition models and a high false positive rate.
By converting the training wireless signal samples into time-spectrum samples, global and local features are extracted using teacher and student identification models. Feature associations are established through cross-dimensional correlation matrices and alignment-free attention weights, and alignment-free feature distillation loss is used for training to obtain a lightweight RF fingerprint recognition model.
It enhances the distinguishability of wireless node identity features, reduces the misjudgment rate in short-frame wireless signal scenarios, improves the consistency of identity category determination, and maintains the model's lightweight deployment advantage.
Smart Images

Figure CN122262833A_ABST