面向鲁棒频谱感知的模糊K-means++算法  

A Fuzzy K-means++Algorithm for Robust Spectrum Sensing

作　　者：任瑾璇 麻淑婉 王永华[1] 万频[1] REN Jinxuan;MA Shuwan;WANG Yonghua;WAN Pin(School of Automation,Guangdong University of Technology,Guangzhou 510006,China)

机构地区：[1]广东工业大学自动化学院,广州510006

出　　处：《电讯技术》2022年第11期1677-1682,共6页Telecommunication Engineering

基　　金：国家自然科学基金资助项目(61971147);广东省研究生教育创新计划项目(2020JGXM040)。

摘　　要：恶意用户通过向数据融合中心发送伪造的频谱感知数据,解决自身频谱资源短缺问题,但会极大地降低频谱感知系统的检测概率。为了解决此问题,提出了基于模糊K-means++的数据融合算法。该算法首先引入模糊处理机制处理样本的数据特征值,以此来增加样本间的差异性;然后将模糊处理后的数据发送到融合中心,融合中心采用离群点挖掘的方法排除恶意用户,并对保留下来的用户进行融合,使样本向量具有鲁棒性;最后运用K-means++算法对样本向量进行聚类。该算法利用轮盘法选择聚类中心,可有效抵御恶意用户的攻击,提高系统感知性能;无需知晓信号与噪声的分布等一些先验信息,也避免了繁杂的门限推导。从仿真结果可以看出,该算法对抵御恶意用户攻击具有突出的效果,有效提升了协同频谱感知系统的稳定性和鲁棒性。Malicious users send forged spectrum sensing data to fusion center to solve their own spectrum resource shortage problem,which greatly reduces the detection probability of spectrum sensing system.In order to solve this problem,this paper proposes a data fusion algorithm based on fuzzy K-means++algorithm.Firstly,fuzzy processing mechanism is proposed to deal with the eigenvalues of sample data,so as to increase the difference between samples.Then,outliers mining is proposed to eliminate malicious users,and fusion center merges together retained users to make sample vector robust.Finally,K-means++algorithm is used to cluster sample vector by roulette wheel selection clustering center,which can effectively resist malicious users attack,and improve system awareness performance.The proposed algorithm does not need prior information,such as signal and noise distribution,and avoids complicated threshold derivation.Simulation results show that the algorithm has outstanding effect on resisting malicious user attacks,and effectively improves the stability and robustness of cooperative spectrum sensing system.

关 键 词：协同频谱感知 恶意用户 模糊处理 K-means++算法 

分 类 号：TN911[电子电信—通信与信息系统]