Search papers, labs, and topics across Lattice.
This paper introduces Polar Monotonic Kolmogorov-Arnold Networks (Polar MKAN), a novel architecture designed for RF fingerprinting that enhances interpretability by ensuring each latent dimension is exclusively influenced by either magnitude or phase. The method significantly improves disentanglement in feature representation, achieving a DCI Disentanglement score of 57.2%, compared to a maximum of 12.9% from traditional unpartitioned approaches. Additionally, the authors assess the model's detection accuracy and its robustness to blind carrier frequency offset compensation using both synthetic and real-world data.
Achieving over four times better feature disentanglement in RF fingerprinting could revolutionize device authentication in security-sensitive applications.
Radio frequency (RF) fingerprinting authenticates wireless devices from hardware-induced I/Q impairments, typically with deep learning feature extractors that are accurate but opaque, limiting their use in security critical settings. We propose Polar Monotonic Kolmogorov-Arnold Networks (Polar MKAN), a block partitioned monotonic encoder on polar inputs in which each latent dimension depends exclusively on magnitude or phase, yielding channel separation and monotone responses by construction. On a synthetic gain and carrier frequency offset (CFO) benchmark, Polar MKAN reaches 57.2 percent DCI Disentanglement versus at most 12.9 percent for unpartitioned baselines. We further evaluate the detection accuracy trade off on real data and the sensitivity to blind CFO compensation.
