Signal Fingerprint-Based Identification for Multi-rotor UAV Remote Control Terminal

In the past decade, with the development of Unmanned Aerial Vehicles (UAV), research about multi-rotor UAV has become one hot spot in its area. Along with the increasing number of UAV-users, illegal-flying occasions appear more frequently, which will affect citizens' daily life. The regulation of safe and ordered UAV-flying has drawn great attentions from the whole society. Recognizing the remote controller individuals is one crucial part for regulating UAV flight. Due to the subtle signal differences between controller individuals, some traditional identification methods are no longer applicable. This paper proposes two fingerprint-based methods for identifying different remote controllers of UAVs and compares the two identification methods. By extracting key features of the constellation diagrams and bispectrum, then applying five machine learning methods, such as decision tree(Tree), Naive Bayes (NB), Discriminant analysis (Discr), k-nearest neighbor (KNN) and Error Correction Output Coding (ECOC), we can achieve accurate UAV controller recognition. By using constellation diagrams as key features,the experiment results show that five machine learning methods can achieve more than 95% accuracy of recognition for four individuals when the SNR is 10dB. By using bispectrum as key features, the decision tree can obtain more than 95% recognition accuracy when the SNR is 5dB, and discr still achieves more than 95% recognition accuracy when the SNR is 0dB, while the other three methods perform in between.