A Novel Lightweight Hybrid Model for Drone Identification Leveraging RF Fingerprints

As drones become more widely used, effectively identifying and managing various types of drones has become an urgent issue that needs to be addressed. Radio frequency fingerprint (RFF) recognition technology can provide a reliable method for identifying drones, helping to prevent the intrusion of unauthorized drones and enhancing the security of airspace management. However, these methods often involve complex models that are difficult to deploy locally. To address this, this paper proposes a lightweight and efficient hybrid model that combines the effective local inductive biases of convolutional neural networks (CNNs) with the global modeling capabilities of transformers, achieving a fusion of local and global features. Additionally, we introduced a spectral attention mechanism to enhance useful information in the signal while suppressing noise. Simulation results indicate that our model consistently achieves high recognition accuracy across various distances, with a peak accuracy of 99.11%. Under low signal-to-noise ratio conditions, our model significantly outperforms other models. For instance, at -5dB, our model achieves an impressive recognition accuracy of 81.25%, compared to MobileNetV2’s 57.14%, a difference of nearly 24 percentage points.