Research On Robust Spectrum Sensing Based On Clustering Algorithm

In view of the practical contradictions of insufficient spectrum resources and the proliferation of spectrum users,as well as malicious users’ attacks on cognitive radio networks,the perception、allocation and application of spectrum resources have become important topics in cognitive radio.At the same time,how to ensure the security of the cognitive network and whether the interference of malicious users can be eliminated or reduced are related to the smooth progress of spectrum allocation and spectrum application.Spectrum sensing technology still has some areas worth optimizing in resisting malicious user attacks.In this paper,some security issues related to spectrum sensing are studied,and effective signal extraction methods and machine learning algorithms are applied to the security issues of spectrum sensing technology,so as to ensure the security of the sensing network.The specific contents are as follows:In order to ensure the security of spectrum sensing,this paper proposes a data processing mechanism based on fuzzy mathematics and K-means++ algorithm to reduce the interference of malicious users.This paper is set in the scenario of multi-antenna parallel fusion.In order to accurately extract the sample data that can reflect the current perception environment and make the perception environment represented by the sample data and weaken the influence of malicious attacks,this paper proposes a fuzzy processing mechanism using the in perceptual samples.Each sensory sample data is given a different membership value,and the K-means++algorithm is used to effectively select the initial cluster center,which avoids the performance loss caused by mistakenly selecting malicious samples as the initial cluster center.The overall design idea of the program are follows:First of all,extract the perceived data received by each antenna of the detection node,and then assign each sample characteristic through the enviter function in the fuzzy processing mechanism.The processed data is trained by the K-Means ++algorithm to obtain a suitable classifier,and finally the classification of the test data set to determine the status of the host user.This scheme does not need to perceive the prior information of the environment,avoids complicated threshold derivation.While ensuring the security of the perception system,it also shows good performance in the case of low signal-to-noise ratio.The K-Means ++ algorithm used in the selection of the initial clustering center,which effectively enhances the accuracy and perceptual efficiency of the perceived results.In order to further resist attacks from malicious users and ensure the security of cognitive radio networks,this paper proposes a robust spectrum sensing scheme based on LU matrix decomposition and ISODATA clustering algorithm.In the multi-antenna system,in order to reduce the interference of malicious users in the network,this scheme proposes a feature extraction method based on LU matrix decomposition and a data fusion method based on outlier mining.The overall design idea of this scheme is as follows: first,perform feature extraction on the sensing data received by each antenna of the detection node,then decompose the sensing matrix by LU matrix,and take the decomposed U matrix as a new sample matrix,Finally calculated a two-dimensional feature vector.In the data fusion stage,the outlier mining operation is performed on this two-dimensional feature vector,malicious users are initially excluded,and data cleaning is performed on them,Finally,the cleaning data is trained by the ISODATA cluster algorithm,and finally obtain the corresponding classifier,using the classifier to determine the status of the main user.This scheme can change the value of the preset parameters according to the current sensing environment and the way of malicious user attack,which can better ensure the security of the cognitive network.It can be seen from the simulation that this scheme can also improve the performance of spectrum sensing.