A Robust and Low-Complexity RF Fingerprinting Method Using Gated Multilayer Perceptron for Cross-Receiver Conditions

Radio Frequency Fingerprinting (RFF) has emerged as a critical technology for the identification and authentication of wireless devices by exploiting the unique hardware-induced characteristics of transmitted signals. In recent years, Deep Learning (DL) has notably enhanced the performance of RFF, but the previously proposed DL-based RFF methods generally suffer from high computational complexity and lack robustness across diverse environments, particularly in cross-receiver scenarios. These limitations hinder their applicability in real-time and resource-constrained systems. To address these challenges, a novel RFF method utilizing a Gated Multilayer Perceptron (gMLP) is proposed in this paper. Specifically, the gMLP integrates gated linear units into the traditional multilayer perceptron. This can enhance information flow management and complex pattern capture at low computational costs, improving performance in cross-receiver scenarios. Experimental results demonstrate that the proposed method achieves an accuracy of 91.8% in cross-receiver scenarios, utilizing only 12,000 parameters and an inference time of 5.5 ms. This performance significantly outperforms several baseline models, including convoultion neural network and ResNet.