Deep Learning for UAV Classification: Impact of Noise and Multipath Fading in RF Signals

The increasing presence of unmanned aerial vehicles (UAVs) raises serious security concerns, particularly regarding unauthorized drone operations. Recent U.S. security statistics report a sharp rise in unauthorized UAV activities, with the Federal Aviation Administration (FAA) receiving over 100 monthly reports of illegal drone operations near airports. In 2024 alone, Dedrone records 1.19 million unauthorized drone flights across major U.S. cities, highlighting the need for robust UAV detection and classification systems. In this work, a lightweight Convolutional Neural Network (CNN) model is proposed for RF-based UAV classification under noisy and multipath fading conditions. The proposed CNN consists of multiple convolutional blocks, max-pooling layers, fully connected dense layers, and dropout regularization to enhance feature extraction and prevent overfitting. We evaluate the model under four different experimental setups to assess its generalization and performance robustness. CNN is evaluated under four different experimental setups to assess its generalization and performance robustness. It achieves an overall accuracy of 99.1% on the clean original dataset and an average cross-validation accuracy of 98.7%, confirming strong generalization. For the proposed CNN model under noisy and faded conditions, the dataset is modified with Additive White Gaussian Noise (AWGN) at -2 dB SNR and Rayleigh fading, with the data split into 80% for training and 20% for testing. In this scenario, the model achieves an accuracy of 90.00%. When trained on noisy data and tested on the original dataset, the accuracy is 87.27%. When training on the original dataset and testing on the noisy, faded dataset, the accuracy drops to 79.00%. The integration of dropout layers and optimized dense configurations strengthens the model’s resilience, making it a promising solution for real-time UAV classification in defense, surveillance, and airspace monitoring applications.