Towards a Unified Few-Shot Learning Evaluation Framework for RF Fingerprinting
Published in ICCCN, 2025
This paper proposes a unified evaluation framework for few-shot RF fingerprinting under realistic domain-shift conditions. We systematically compare three transfer learning approaches across multiple datasets and environmental scenarios (e.g., transmitter location, distance, and device configuration) to analyze robustness and performance degradation. Our results identify the most challenging deployment settings and provide insights into how environmental factors affect cross-domain generalization in RF-based device identification.
Recommended citation: S. Shi, V. Mahzoon, X. Wang, S. Mao, J. Wu and S. Vucetic, “Towards a Unified Few-Shot Learning Evaluation Framework for RF Fingerprinting,” 2025 34th International Conference on Computer Communications and Networks (ICCCN), Tokyo, Japan, 2025, pp. 1-9, doi: 10.1109/ICCCN65249.2025.11133814},