网络伪装隐形文本特征检测及数据挖掘方法  

Method of Data Mining and Hidden Text Data Feature Detection for Network Intrusion

作　　者：康凤[1] 蒋小惠[1] 冯梅[1] 

机构地区：[1]成都航空职业技术学院工程实训中心,成都610100

出　　处：《科技通报》2014年第4期113-115,共3页Bulletin of Science and Technology

摘　　要：基于特征分解原理,提出一种多维空间协方差矩阵数据挖掘算法,进行了最优化特征检测性能迭代和子空间文本数据特征检测算法的设计研究。提出采用K-L变换的特征压缩器设计进行高维特征向量的特征压缩,提高算法精度和减少计算量。在子空间中将文本数据空间分解为两个空间向量,采用两个空间向量的正交特性进行降噪去伪处理和特征量的检测和提取。仿真实验对高度伪装隐形文本入侵特征检测,采用了DARPA数据库作为实验数据为研究对象,实验表明新算法能有效检测出信号出现的两个峰值,检测效果明显,检测性能较高,具有良好的入侵文本特征数据挖掘性能。An improved hyperspace covariance matrix data mining algorithm was proposed, the optimum feature detectionperformance iteration algorithm and the subspace text data feature detection algorithm were designed in the same time. Forreducing the dimensions of extracted feature, the feature compressor was designed based on K-L transform for the benefitof improving the accuracy and reducing the computing load. The subspace was divided into two spaces vectors, and the orthonormal property of the two vectors was taken good use for reducing the noise and eliminating the camouflage. Simulationexperiment was worked with the DARPA database for the detection and data mining. Simulation result shows that the proposed method can detect the signal as the showing of two peaks, the detection performance is various, the intrusion data ismined effectively.

关 键 词：网络入侵 检测 特征提取 数据挖掘 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]