【作者】 王欣；

【导师】 王卫东；

【作者基本信息】 北京邮电大学 ， 电子科学与技术， 2018， 硕士

【摘要】 近年来,无线通信的安全问题逐渐成为了各界关注和重视的热点话题。在通信网络安全领域,识别不合法设备是一个十分重要的问题。射频指纹识别技术因其难以通过软件模仿,同时唯一标识设备的特性而得到人们广泛的研究。在设备生产过程中,由于制造工艺限制,即使是同一厂家生产的同一型号的设备会由于材料成分占比等方面的差异而呈现不同的特征,就像每个人都有不同的指纹一样。这些物理层的特征称为射频指纹特征。同时,这些特征可以在通信信号中有所体现,即可以通过分析接收到的射频信号来得到射频指纹特征。因此,通过对射频指纹特征进行提取和分类来分辨入侵设备和授权设备,这在一定程度上可以提高通信系统的安全性。目前,射频指纹识别技术主要由基于瞬态信号和稳态信号两种方法实现。考虑到稳态信号易提取、稳定性高的特点,本文主要讨论基于稳态信号提取射频指纹特征的方法来区分射频设备。主要的研究结果如下:本文通过介绍射频指纹产生机理和特点,建立射频指纹识别系统模型。同时,结合双谱在理论上可以抑制加性高斯白噪声的特性,针对信号双谱特征展开研究。改进了传统的基于局部积分双谱的特征提取方法,提出了采用双谱能量熵及颜色矩作为射频指纹特征,以支持向量机作为分类器的射频指纹识别方法BEECM(Bispectrum-based method using energy entropy and color moments)。通过仿真证明 了本算法的有效性,尤其是在低信噪比(SNR,Signal to Noise Ratio)条件下。另外,通过与基于矩形积分双谱波形熵和能量熵的方法(EWEB,Energy entropy and waveform energy entropy of SIB)进行对比,发现本算法在分类准确度上有一定的优势,当信噪比SNR=0dB时,正确率接近80%,随着SNR的升高,其分类准确度随之提高,在20dB时达到95%左右。另外,本文利用软件无线电开发平台GNU Radio和同一厂家生产的同一型号的4台通用软件无线电外设(USRP,Universal Software Radio Peripheral)组成软件无线电系统完成实际信号的收发。其中3台通用软件无线电外设作为信号发送设备,1台通用软件无线电外设作为信号接收设备。通过对实际接收的信号进行处理,分别提取两种方法中用到的射频指纹特征,并通过训练和测试得到相应的分类准确度。实验结果进一步证明了本文提出的BEECM方法的优势和可行性。更多还原

【Abstract】 The security of wireless communication is a matter of widespread concern.In the field of communication network security,identifying illegal devices is a key problem.Radio frequency fingerprinting gets a widespread of research,because radio frequency fingerprinting feature is difficult to imitate by software,while it uniquely represent the devices.It has been proved that,in the process of equipment production,even the same type of equipment produced by the same manufacturer will have different characteristics due to the difference in the proportion of material components,just like everyone has a different fingerprint.These physical layer characteristics,called radio frequency fingerprinting features,will be reflected in communication signals.These features are unique and can not be forged.They can be extracted by analyzing the received radio frequency signals.Therefore,by recognizing the authorized devices from the invasion devices through extracting and classifying radio frequency fingerprinting features,the security and privacy of the communication can be enhanced.In this paper,we focus on steady state signals and extract the radio frequency fingerprinting features to identify radio frequency devices.The main research results are shown as follows:Considering the mechanism and characteristics of radio frequency fingerprinting,the radio frequency fingerprinting identification system model is established.In this paper,a novel method extracting the energy entropy and color moment of bispectrum as radio frequency fingerprinting features,employing the support vector machine(SVM)as the classifier is proposed.It is called BEECM(Bispectrum-based method using energy entropy and color moments).The simulate results prove the effectiveness of the proposed BEECM method in theoretical,especially when SNR(Signal to Noise Ratio)is low.In addition,it shows that the method has a certain advantage in the identification accuracy.When SNR= OdB,the correct rate is nearly 80%.With the increase of SNR,the accuracy of classification increased,achieving above 95%at 20dB.What’s more,in this paper,software radio system consists of software radio development platform GNU Radio and the 4 USRPs from the same manufacturer of same type are employed as the actual signal transceiver.3 USRPs are acted as a signal sending devices,1 USRP is used as a signal receiving device.Through the processing of the received signals,the radio frequency fingerprinting features are extracted respectively.Through training and testing,we get the classification accuracy.The experimental results further demonstrate the advantages and feasibility of the BEECM method proposed in this paper.更多还原