Research On Wireless Device Identification Based On Device Fingerprint

With the advent of the 5G era and the prosperity of Internet of Things(IoT)applications,communication between people,people and things,things and things are getting more and more popular,and wireless networks are playing an increasingly important role.The open nature of the wireless communication medium enables malicious users to interfere and wiretap the communication between legitimate users by disguising their identity,which seriously affects the security of wireless network communication.The traditional identity authentication mechanism based on encryption is facing more and more challenges in the new IoT application environment.The physical layer identity authentication technology has received more and more attention in recent years due to its strong anti-spoofing-attack capability.This thesis studies the use of physical layer characteristics to construct unique fingerprints of wireless devices,which is then used to identify specific wireless devices.This thesis proposes a device fingerprint extraction method based on multi-differential interval constellation trace figure.This method determines the rotation angle in the constellation of the differential results of the received signal.As the phase rotation is resulted by the inherent carrier frequency offset of the transmitter and is unique with respect to the specific transmitter,it can be used for identification of the signal transmitter.Specifically,by performing a differential operation on the received signal,the carrier frequency can be converted to a phase rotation in the constellation figure,thus different levels of carrier frequency offset can be visually indicated.The angle of phase rotation is related to the differential interval.In differential operations,the use of shorter differential interval can be applied to a wider range of carrier frequency offsets without phase confusion.Longer differential interval can differentiate smaller carrier frequency differences and reach high frequency resolution.In this thesis,we propose to combine three different differential intervals,namely short,medium and relatively long differential intervals as device fingerprint,and solve the problem that the traditional single differential interval constellation trace figure cannot balance the frequency offset resolution and frequency offset coverage.The experimental results show that the accuracy of the device fingerprint using the multi-differential interval constellation trace figure is better than the traditional method.Especially in the case of low signal-to-noise ratio,the recognition accuracy of device fingerprints based on multi-differential interval is 2%higher than that of traditional single differential interval device fingerprints.Further,this paper proposes a wireless device identification algorithm based on generative adversarial network.Different from previous studies where the identification of devices is modeled as classification problems,this thesis proposes an algorithm that can detects whether the device under test is a legitimate user device and identify the identity of it.The main idea behind the algorithm is to use the adversarial training to learn the optimal representation of the latent space of the legitimate user and its distribution in the case where only the legitimate user's signal samples are known in the training,and re-encode the optimal representation by the encoder.When the signal under tested is from a legitimate user,the representation of the latent space is similar to its reconstruction vector;and when the signal under tested is from an illegal user,the representation of the latent space is significantly different from its reconstruction vector.This is used as the basis to determine whether the signal under tested is from a legitimate user.The optimal representation of the latent space is further used as the characteristics in the identification of legitimate user.The experimental results show that the proposed device identification algorithm based on generative adversarial network can accurately discover illegal users and accurately retrieve the identity of legitimate users.It is also shown that the effect of illegal user detection and legitimate user identification using the multi-differential interval constellation trace figure is better than that of the traditional single differential interval constellation trace figure.