大数据驱动下复杂网络输入节点在线筛选仿真
|
摘要
目前网络输入节点筛选方法漏检率和运行复杂度高,提出基于度量阈值的大数据驱动下复杂网络输入节点在线筛选方法。将数据样本属性的数量、数据样本的数量和预设信息熵阈值当作节点数据降维的输入,同时计算输入节点中各数据属性信息熵,将计算出的信息熵与预设信息熵阈值进行对比,以此得到节点数据样本中心化函数和协方差函数,根据所得结果计算选定变换基,将选定变换基代入节点数据降维,获得节点数据降维结果。将节点数据降维结果引入复杂网络输入节点筛选中,通过数据采样对任意节点自适应度量阈值进行计算,利用所得阈值对某节点是否为恶意节点进行判断。将各簇头的节点当作中心节点,利用数据样本分别和各簇头节点实行聚类处理,并对大数据驱动下复杂网络中有恶意行为的所有输入节点实行属性划分操作,利用恶意节点的筛选,对聚类之后的网络输入节点再次筛选,直到选定的网络输入节点全部被筛选完毕。实验结果表明,所提方法筛选漏检率和运行复杂度均较低,具有较强的可靠性。
At present, the network input node screening method has high missed detection rate and high running complexity. An online filtering method for complex network input nodes driven by big data based on metric threshold is proposed. The number of data sample attributes, the number of data samples and the preset information entropy threshold are taken as the input of the node data dimension reduction. At the same time, the entropy of each data attribute in the input node is calculated, and the calculated information entropy and preset information entropy threshold are performed. By comparison, the node data sample centering function and covariance function are obtained. Based on the obtained results, the selected transformation base is calculated, and the selected transformation base is substituted into the node data dimension reduction to obtain the dimensionality reduction result of the node data. The dimensionality reduction results of the node data are introduced into the complex network input node filter. The adaptive metric threshold of any node is calculated through data sampling, and the resulting threshold is used to judge whether a certain node is a malicious node. The nodes of each cluster head are treated as a central node, and the data samples are used to perform clustering processing on each cluster head node respectively, and attribute partition operations are performed on all input nodes that have malicious behavior in a complex network driven by big data, and the malicious nodes are used. Filtering, filtering the network input nodes after clustering until the selected network input nodes are all screened. The experimental results show that the proposed method has a low screening rate and a low complexity of operation, and has a strong reliability.
At present, the network input node screening method has high missed detection rate and high running complexity. An online filtering method for complex network input nodes driven by big data based on metric threshold is proposed. The number of data sample attributes, the number of data samples and the preset information entropy threshold are taken as the input of the node data dimension reduction. At the same time, the entropy of each data attribute in the input node is calculated, and the calculated information entropy and preset information entropy threshold are performed. By comparison, the node data sample centering function and covariance function are obtained. Based on the obtained results, the selected transformation base is calculated, and the selected transformation base is substituted into the node data dimension reduction to obtain the dimensionality reduction result of the node data. The dimensionality reduction results of the node data are introduced into the complex network input node filter. The adaptive metric threshold of any node is calculated through data sampling, and the resulting threshold is used to judge whether a certain node is a malicious node. The nodes of each cluster head are treated as a central node, and the data samples are used to perform clustering processing on each cluster head node respectively, and attribute partition operations are performed on all input nodes that have malicious behavior in a complex network driven by big data, and the malicious nodes are used. Filtering, filtering the network input nodes after clustering until the selected network input nodes are all screened. The experimental results show that the proposed method has a low screening rate and a low complexity of operation, and has a strong reliability.
引文
[1] 马润年,王班,王刚,等.基于互信息的通信网络节点重要性度量方法[J].电子学报,2017,45(3):747-752.
[2] 张成军,刘超,郭强.大数据网络环境下异常节点数据定位方法仿真[J].计算机仿真,2017,34(5):273-276.
[3] 杜景林,侯大俊.参与感知网络中转发节点抽样优化算法[J].计算机工程与设计,2016,37(10):2597-2601.
[4] 吴大鹏,司书山,闫俊杰,等.基于行为特征分析的社交网络女巫节点检测机制[J].电子与信息学报,2017,39(9):2089-2096.
[5] 占清华.网络入侵后最优逃避攻击节点的选取与研究[J].现代电子技术,2017,40(20):77-79.
[6] 顾亦然,朱梓嫣.基于LeaderRank和节点相似度的复杂网络重要节点排序算法[J].电子科技大学学报,2017,46(2):441-448.
[7] 刘保见,李青,张效义.一种提高网络寿命与节点定位性能的WSNs分群算法[J].计算机应用研究,2016,33(11):3417-3421.
[8] 柳治谱,王勇.基于无线传感器网络的坡面异常信号检测方法研究[J].传感技术学报,2017,30(10):1536-1541.
[9] 易俗,殷慧文,王闯,等.多因素自适应心跳检测算法研究[J].计算机工程与应用,2017,53(24):86-93.
[10] 陶偲,姚飞,刘守印.基于相位差测距的WSN节点测距数据滤波和定位算法的研究[J].电子设计工程,2017,25(7):35-40.
[2] 张成军,刘超,郭强.大数据网络环境下异常节点数据定位方法仿真[J].计算机仿真,2017,34(5):273-276.
[3] 杜景林,侯大俊.参与感知网络中转发节点抽样优化算法[J].计算机工程与设计,2016,37(10):2597-2601.
[4] 吴大鹏,司书山,闫俊杰,等.基于行为特征分析的社交网络女巫节点检测机制[J].电子与信息学报,2017,39(9):2089-2096.
[5] 占清华.网络入侵后最优逃避攻击节点的选取与研究[J].现代电子技术,2017,40(20):77-79.
[6] 顾亦然,朱梓嫣.基于LeaderRank和节点相似度的复杂网络重要节点排序算法[J].电子科技大学学报,2017,46(2):441-448.
[7] 刘保见,李青,张效义.一种提高网络寿命与节点定位性能的WSNs分群算法[J].计算机应用研究,2016,33(11):3417-3421.
[8] 柳治谱,王勇.基于无线传感器网络的坡面异常信号检测方法研究[J].传感技术学报,2017,30(10):1536-1541.
[9] 易俗,殷慧文,王闯,等.多因素自适应心跳检测算法研究[J].计算机工程与应用,2017,53(24):86-93.
[10] 陶偲,姚飞,刘守印.基于相位差测距的WSN节点测距数据滤波和定位算法的研究[J].电子设计工程,2017,25(7):35-40.