基于特征聚类的网络恶意干扰信号协同过滤  

Collaborative filtering of network malicious interference signals based on feature clustering

作　　者：梁超 付明林 LIANG Chao;FU Minglin(Henan Vocational College of Water Conservancy and Environment,Zhengzhou 450011,China)

机构地区：[1]河南水利与环境职业学院,郑州450011

出　　处：《自动化与仪器仪表》2024年第7期322-325,330,共5页Automation & Instrumentation

基　　金：河南省科技攻关计划(232102320009)。

摘　　要：为解决网络恶意干扰信号引发的网络安全问题,提出基于特征聚类的网络恶意干扰信号协同过滤方法。基于网络恶意干扰信号特点,将包络起伏度、时域矩偏度以及时域矩峰度三者协同作为描述网络恶意干扰信号的特征,采用FCM特征聚类算法,根据每个网络恶意干扰信号特征样本点到聚类中心的欧式距离,将具有相似特征的信号样本点归为同一聚类簇,完成网络恶意干扰信号特征聚类;将特征聚类结果利用网络信号矩阵投影至特征子空间,并依据主成分分析法获取主分量特征值划分子空间为正常信号和恶意干扰信号子空间,将恶意干扰信号子空间残差信号平方预测误差与恶意干扰信号过滤阈值比较,判定恶意干扰信号并将其过滤。实验结果表明,该方法可有效过滤掉网络恶意干扰信号,得到了较为纯净网络信号,且过滤准确率高,漏判和误判的情况较少。To address the network security issues caused by malicious interference signals,a collaborative filtering method for network malicious interference signals based on feature clustering is proposed.Based on the characteristics of network malicious interference signals,the combination of envelope undulation,time-domain moment skewness,and time-domain moment kurtosis is used to describe the characteristics of network malicious interference signals.The FCM feature clustering algorithm is used to classify signal sample points with similar features into the same cluster according to the Euclidean distance from each network malicious interference signal feature point to the cluster center,completing the clustering of network malicious interference signal features;The feature clustering results are projected onto the feature subspace using the network signal matrix,and the principal component feature values are obtained using principal component analysis to divide the subspace into normal signal and malicious interference signal subspaces.The residual signal squared prediction error in the malicious interference signal subspace is compared with the filtering threshold of the malicious interference signal,and the malicious interference signal is determined and filtered.The experimental results show that this method can effectively filter out malicious network interference signals,obtain relatively pure network signals,and have high filtering accuracy,with fewer cases of missed and false positives.

关 键 词：特征聚类 恶意干扰信号 协同过滤 FCM聚类算法 过滤阈值 残差信号 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]