Abstract
The proliferation of wireless devices has intensified concerns over network security. Radio Frequency Fingerprinting (RFF) has emerged as a promising technique for passive device authentication by leveraging unique hardware-induced waveform signatures. Recent studies have demonstrated that graphical deep learning approaches, particularly those employing Convolutional Neural Networks (CNNs), can achieve up to 98% classification accuracy using time-frequency domain representations of real-world signals. However, deploying such models on edge devices remains challenging due to limitations in power, size, and computational capacity, especially when incorporating preprocessing steps like the Short-Time Fourier Transform (STFT).
This paper presents a complete System-on-Chip (SoC) implementation of a graphical deep learning-based RFF system using the hls4ml framework. Targeting the AMD RFSoC XCZU28DR platform, the proposed solution integrates STFT and CNN inference within a unified hardware architecture. We evaluate three deployment strategies - hardware-only, software-only, and hybrid hardware-software co-implementation - to assess their suitability for real-time RFF in edge computing scenarios. Experimental results demonstrate the feasibility of efficient and scalable device authentication, bridging the gap between algorithmic performance and practical deployment.
This paper presents a complete System-on-Chip (SoC) implementation of a graphical deep learning-based RFF system using the hls4ml framework. Targeting the AMD RFSoC XCZU28DR platform, the proposed solution integrates STFT and CNN inference within a unified hardware architecture. We evaluate three deployment strategies - hardware-only, software-only, and hybrid hardware-software co-implementation - to assess their suitability for real-time RFF in edge computing scenarios. Experimental results demonstrate the feasibility of efficient and scalable device authentication, bridging the gap between algorithmic performance and practical deployment.
| Original language | English |
|---|---|
| Title of host publication | 2025 IEEE Asia Pacific Conference on Circuits and Systems (APCCAS) |
| Place of Publication | Piscataway, NJ |
| Publisher | Institute of Electrical and Electronics Engineers (IEEE) |
| Number of pages | 5 |
| ISBN (Electronic) | 9798331589073 |
| ISBN (Print) | 9798331589080 |
| DOIs | |
| Publication status | Published - 2025 |
| Event | IEEE Asia Pacific Conference on Circuits and Systems (21st : IEEE) - Busan, Korea, Republic of Duration: 12 Oct 2025 → 15 Oct 2025 Conference number: 21st |
Publication series
| Name | |
|---|---|
| ISSN (Print) | 2837-4576 |
| ISSN (Electronic) | 2768-3516 |
Conference
| Conference | IEEE Asia Pacific Conference on Circuits and Systems (21st : IEEE) |
|---|---|
| Abbreviated title | APCCAS 2025 |
| Country/Territory | Korea, Republic of |
| City | Busan |
| Period | 12/10/25 → 15/10/25 |
Fingerprint
Dive into the research topics of 'System-on-chip implementation of deep learning RF fingerprinting for device authentication'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver