Enhanced UAV Classification: Gaining Deeper Insights from Mechanical Vibration Over RF Characteristics

UAV classification plays a critical role in enhancing authentication and authorization within UAV-based aerial networks, providing a robust mechanism to prevent unauthorized access and ensure secure operations. This paper presents a novel approach to UAV classification by combining RF characteristics and mechanical vibration data, exploring how these methods can strengthen the security and reliability of UAV communication systems. Vibration data, captured through onboard accelerometers, is investigated as a complement to traditional RF fingerprinting, offering an alternative for UAV identification in dynamic environments where RF-based methods may face limitations. Deep learning models, including a Recurrent Neural Network (RNN) for vibrational data and a Convolutional Neural Network (CNN) for 5G communication data, were developed to classify UAVs with high accuracy. The results demonstrate that integrating mechanical vibration data significantly enhances classification performance compared to using RF fingerprinting alone, providing new avenues for secure UAV authentication and authorization. Furthermore, this study emphasizes the importance of efficient, real-time classification methods for deployment in 5G networks, ensuring both the security and operational integrity of UAV systems.