Radio fingerprint-based UAV detection and identification using discrete wavelet feature extraction and deep learning approaches

The increasing use of drones in various industries presents security detection challenges due to their high altitude and remotely controlled capabilities. Researchers have developed methods for identifying unmanned aerial vehicles (UAVs), including camera, audio, radar, and thermal-based techniques. Radio fingerprinting is the effective method for detecting drones at high altitudes. This research used discrete wavelet transform (DWT), machine learning (ML) and deep learning (DL) models for UAV identification and detection. The Random Forest feature selection method improved classification models' accuracy and reduced computational time. The long-short term memory (LSTM) model demonstrated promising results, achieving classification accuracy of 95.27%, 87.54%, 82.21%, and 96.7% for identifying UAVs, controller signals, UAV model specifications, and mode of operation. It also demonstrated accuracy of 86.42% and 88.79% in non-line of sight. This research offers valuable insights into practical methods for identifying and detecting UAVs, with significant commercial implications.