Designing an attention-based approach for RF fingerprinting in drone detection and classification

Area of Science:

• * Electrical Engineering
• * Artificial Intelligence
• * Signal Processing

Background:

• * The proliferation of unmanned aerial vehicles (UAVs) necessitates advanced radio frequency (RF) fingerprinting systems for detection, identification, and operational mode classification.
• * Existing methods often lack the computational efficiency and accuracy required for real-world UAV monitoring.

Purpose of the Study:

• * To introduce a computationally efficient transformer-based attention architecture for analyzing RF signals from UAVs.
• * To develop a robust system capable of detecting, classifying UAVs, and identifying their flight modes in diverse conditions.

Main Methods:

• * A two-stage feature extraction pipeline involving correlation filtering and ANOVA feature selection.
• * A multi-headed self-attention encoder to process spectral sequences of RF signals.
• * Evaluation on the DroneRF and VTI_DroneSET_FFT datasets, including single- and multi-drone scenarios across 2.4 GHz and 5.8 GHz bands.

Main Results:

• * Achieved high cross-validation accuracy: 100.00% for detection, 99.80% for classification, and 99.23% for flight mode identification on the DroneRF dataset.
• * Demonstrated strong performance on the challenging VTI_DroneSET_FFT dataset, reaching 88.10% accuracy in the complex multi-drone 5.8 GHz scenario.
• * The model's low parameter count and FLOPs make it suitable for edge deployment.

Conclusions:

• * The proposed transformer-based architecture offers an intelligent and efficient solution for UAV RF fingerprinting.
• * The system demonstrates high accuracy and robustness in real-world applications, outperforming baseline methods.
• * Attention analysis confirms the effectiveness of the model in identifying and characterizing diverse UAV RF signals.