Abstract: With the development of communication technology, a large number of wireless communication devices are connected to the network. However, due to the openness of wireless networks, malicious users can pretend to be legitimate users to access the network by disguising their identities, which seriously threatens the security of wireless communication networks. Because of the stability and uniqueness of the characteristics of the physical layer of the transmitter, extracting the characteristics of the physical layer of the transmitter as the terminal's identity to identify the device has become a research focus in recent years. This paper analyzes the research progress of device identification based on the physical layer characteristics of the device in recent years. Aiming at the risks of existing fingerprint identification methods, this paper proposes a LoRa device identification model based on device fingerprints, using OneClassSVM single classification algorithm, the identification of illegal equipment is transformed into an abnormal detection problem, the authentication of the legality of the equipment under test is realized, and a higher recognition rate is obtained. A false alarm elimination algorithm is designed to reduce the false alarm rate and verify the randomness. And use random forest, support vector machine (SVM), KNN as classifiers to verify its performance in device identification.

Key words: LoRa, machine learning, device identification, SVM, KNN, random forest

摘要： 随着通信技术的发展，大量的无线通信设备接入网络，然而，由于无线网络的开放性，使得恶意用户可能通过伪装身份冒充合法用户接入网络，严重威胁了无线通信网络的安全．由于发射机物理层特征具有稳定性和唯一性等特性，提取发射机物理层特征作为终端的身份标识对设备进行身份识别成为了近几年研究的热点．本文分析了近年来基于设备物理层特征实现设备身份识别的研究进展，针对现有LoRa设备指纹身份识别方法存在的风险，本文提出了一种基于设备指纹的LoRa设备识别模型，采用OneClassSVM单分类算法，将非法设备识别转化为异常检测问题，实现了对待测设备合法性的鉴权，获得了较高的识别率，设计了一种误报销除算法实现了误警率的降低，并验证了随机森林、支持向量机（SVM）、KNN作为分类器在设备身份识别中的性能．

关键词: LoRa, 机器学习, 设备身份识别, SVM, KNN, 随机森林