Self-Supervised Learning for Radio Frequency Fingerprinting
Dr. Yuki Tanaka, Prof. Masahiro Sato
NTT Docomo Research
Abstract
We propose a self-supervised learning framework for RF fingerprinting that eliminates the need for labeled training data. Using contrastive learning on raw IQ samples, our approach learns device-specific features from unlabeled radio emissions. The method achieves 97% device identification accuracy on a dataset of 100 devices, matching supervised approaches while reducing data labeling effort by 99%.
AI Summary
- Self-supervised contrastive learning eliminates labeling for RF fingerprinting.
- 97% device identification accuracy matching supervised methods.
- Validated on 100 devices across different environments and conditions.
- Reduces data preparation effort by 99%.
Key Findings
- 1Contrastive learning on raw IQ data captures hardware impairment signatures effectively.
- 2Learned representations transfer across different wireless environments.
- 3Robust to varying SNR conditions and channel effects.
Industry Implications
Enables scalable IoT device authentication for 6G massive connectivity scenarios.
Reduces deployment cost of physical layer security solutions.
Applicable to spectrum enforcement and unauthorized transmitter detection.
Read the Original Paper
Access the full paper on arXiv for complete methodology, results, and references.
Open on arXivRelated Papers
Transformer-Based Channel Estimation for Massive MIMO Systems
Tsinghua University — 12 citations
AI/ML PapersFederated Reinforcement Learning for Distributed Network Optimization
Stanford University — 8 citations
AI/ML PapersNeural Architecture Search for Efficient Edge AI in Wireless Networks
Samsung AI Center Seoul — 5 citations