Robust Radio Frequency Fingerprinting With Signal Denoising and Stacked Multivariate Ensemble Learning for Secure Wireless Communications

Radio Frequency Fingerprinting (RFF) has gained significant attention in wireless communication and security research due to its potential for device authentication and intrusion detection. While deep learning-based approaches have shown promising results, existing methods suffer from critical limitations: high computational complexity hinders real-time deployment on resource-constrained hardware, poor robustness under low Signal-to-Noise Ratio (SNR) conditions, and inadequate generalization across different datasets. To address these gaps, this paper proposes a novel and efficient RFF framework that integrates signal denoising preprocessing using Savitzky-Golay Filtering (SGF) with Stacked Multivariate Ensemble Learning (SMvEL). The proposed architecture employs lightweight, homogeneous Convolutional Neural Networks (CNNs) optimized for rapid model training and fast inference, ensuring computational efficiency without sacrificing accuracy. Experimental results on real-world walkie-talkie datasets, as well as two open-source benchmark datasets for cellphones and drones, demonstrate that the proposed method outperforms state-of-the-art deep learning approaches in both accuracy and robustness.