RF-based Identification Framework against Unauthorized UAV Networking in Low-Altitude Economy

Abstract

As unmanned aerial vehicles (UAV) become increasingly prevalent for the low-altitude economy, security concerns are raised such as unauthorized access and control, collisions and interference with manned aircraft, and illegal eavesdropping. Therefore, ensuring secure UAV networking holds significant implications over low-altitude wireless networks. In this paper, we first conduct a comprehensive review focusing on the security challenges and solutions of low-altitude UAV networks. Then, we introduce a radio frequency (RF)-based UAV identification framework to accurately detect and identify different UAVs, which is significant for supporting secure access and authentication in UAV networks. Specifically, we utilize short-time fourier transform (STFT) to extract two-dimensional features from the raw signals, which contain both time-domain and frequency-domain information and can be used to recognize the type and flight mode of UAVs. With the extracted two-dimensional features, we employ a convolutional neural network built with depth wise separable convolution (DSC) to achieve multi-class classifications. Based on a real drone dataset and comparisons against baselines, our experiments show that the proposed DSC-STFT can achieve higher accuracy in different classification tasks with lower computational complexity. Additionally, it also holds good robustness on different types and signal-noise ratio (SNR) levels of noisy data.