大数据网络入侵过程的痕迹数据监测方法研究

发布时间：2018-06-24 23:30

本文选题：大数据网络 + 入侵过程　；参考：《科学技术与工程》2016年14期

【摘要】：大数据网络数据规模巨大,对入侵过程痕迹数据进行监测的效率通常较低,一些带有入侵痕迹的数据特征在大数据环境下,特征逐渐淡化,当前方法无法在淡化的情况下准确采集痕迹数据的特点,无法形成待监测数据与痕迹数据之间的关系,导致监测效率和精度低下。提出一种基于模糊聚类概率的大数据网络入侵过程的痕迹数据监测方法,将采集的痕迹数据转换成频域信号,对其进行频谱或功率谱分析,依据时间变化的幅值将其转换成随频率变化的功率。采用核主元分析对痕迹数据信号特征进行提取,利用非线性转换将样本痕迹数据信号从输入空间映射至高维特征空间,在高维特征空间中通过PCA进行痕迹数据信号的频域特征提取。构建一个数学模型对特征模糊聚类概率进行描述,对待监测数据和痕迹数据之间的特征模糊聚类概率进行计算,通过衡量理论进行对比分析,使大数据网络入侵过程中的痕迹数据被完整的监测。实验结果表明,所提方法不仅所需时间少,而且监测精度高。
[Abstract]:Because of the large scale of big data network data, the efficiency of monitoring intrusion trace data is usually low. Some data features with intrusion trace are gradually desalinated in big data environment. The current method can not accurately collect trace data under the condition of desalination, and can not form the relationship between the monitoring data and trace data, which leads to the low efficiency and precision of monitoring. A method of trace data monitoring in big data network intrusion process based on fuzzy clustering probability is proposed. The trace data collected is converted into frequency domain signal, and the spectrum or power spectrum is analyzed. It is converted to power varying with frequency according to the amplitude of time variation. The feature of trace data signal is extracted by kernel principal component analysis, and the sample trace data signal is mapped from input space to high dimensional feature space by nonlinear transformation. The feature extraction of trace data signal in frequency domain is carried out by PCA in high dimensional feature space. A mathematical model is constructed to describe the feature fuzzy clustering probability and to calculate the feature fuzzy clustering probability between the monitoring data and trace data. The trace data in the process of big data network intrusion is monitored completely. The experimental results show that the proposed method not only requires less time, but also has high monitoring accuracy.
【作者单位】：重庆市气象局;
【分类号】：TP393.08

【相似文献】