Research On Radio Frequency Fingerprint Identification Algorithm Based On Channel Fingerprint Separation For 5G

With the increasing popularity of mobile communication devices and booming development of Internet techniques,wireless communication,playing an important role both in military and civilian terms,has been an integral part of modern society.Recently,with the rapid rise of 5G,people’s life is entering a new era of wireless communication.Nowadays,mobile network is becoming heterogeneous and diversified,requiring the functions of supporting massive devices,low latency and large bandwidth.Wireless network,in comparison with traditional wired network,could easily received massive malicious attacks because of its open communication environment.In 5G communication,there are large quantities of access devices and the channel environment is more complex and changeable.The challenge faced by new era of 5G application is the security problems caused by highly complex communication environment.Traditional protection mechanism in physical layer is based on bits,while security protocols of upper layers focus on the integrity and confidentiality of data to provide the identity authentication of communication parties.Whereas,these security mechanisms usually have security bugs.Therefore,this thesis is to extract and identify features of radio frequency fingerprints based on separation of channel characteristics in the 5G scenario.The main researches,to achieve high accuracy rate of fingerprints identification in complex and changeable channels,are as follows.Firstly,based on the reasons of the generation of different radio frequency fingerprints in transmitters when transmitting,this thesis analyses the specific performances of different radio frequency fingerprints on transmit signals and received signals.Then,this thesis also briefly introduces physical layer design in 5G,and focuses on PRACH signal physical layer design,such as: frame structure and time frequency domain resources.Secondly,due to the flexible mobile communication channel,the device’s unique features extracted from received signals will mix with channel characteristics,this thesis researches the solution of eliminating channel characteristics from radio frequency fingerprints,proposes the RFF estimation algorithm based on channel estimation.This algorithm obtains the received signals which do not contain channel characteristics by channel equalization.Then,this method extracts estimated transmitters’ RFF from received signals,and the differences between different transmitters’ estimated fingerprints will be shown in 5G through simulations.Then,using machine learning classification algorithm to identify the estimated fingerprints and calculate the identity accuracy rate in 5G five channel models.In addition,this algorithm performs well when channel characteristics change.At last,because of the ideal transmit signal sequence and the complex computation.This thesis researches the RFF identification algorithm based on non-channel estimation,eliminating channel characteristics from transmitters’ fingerprints features by obtaining the received signal spectrum,proposes ASo Q and Do Lo S algorithm which utilize noise elimination and logarithmic difference respectively to obtain pure fingerprints without channel characteristics,and also shows sensitive level to channel changing on both two methods.And also,it shows differences of different devices’ fingerprints.In addition,this thesis classifies and identifies two features of different devices in different channels and calculates the identification accuracy rates in 5G.It is easy to see that these two features,with the capacity to identify different transmitters,lead to 80% identity accuracy rate,meanwhile there are high accuracy rates while channel characteristics change.