CAJ | 학술논문

在对波段性网络入侵差异化特征进行提取的过程中，会出现入侵差异化特征伪装程度逐渐升高的情况，导致传统的基于敏感性数据挖掘的波段性网络入侵差异化特征提取方法，由于不能有效区分入侵特征与正常特征，无法有效实现波段性网络入侵差异化特征的有效提取，提出一种基于支持向量机的波段性网络入侵差异化特征提取模型，获取不确定入侵中波段性网络节点的差异化特征以及入侵节点，给出两种不确定入侵中波段性网络节点的差异化特征训练数据样本集，通过非线性映射将数据样本集从原空间映射到高维特征空间中，得到高维特征空间中最优线性分类面，采用支持向量机求解该分类面的优化解，使用网格搜索法，通过调整错分惩罚因子与核宽度，分别训练不同的支持向量机，获取泛化能力最强的参数组合，完成入侵中波段性网络节点的差异化特征所对应的数据集的核参数优化和分类，实现波段性网络入侵节点的差异化特征的有效提取。仿真实验结果表明，所提方法具有很高的准确性及有效性。
재대파단성망락입침차이화특정진행제취적과정중，회출현입침차이화특정위장정도축점승고적정황，도치전통적기우민감성수거알굴적파단성망락입침차이화특정제취방법，유우불능유효구분입침특정여정상특정，무법유효실현파단성망락입침차이화특정적유효제취，제출일충기우지지향량궤적파단성망락입침차이화특정제취모형，획취불학정입침중파단성망락절점적차이화특정이급입침절점，급출량충불학정입침중파단성망락절점적차이화특정훈련수거양본집，통과비선성영사장수거양본집종원공간영사도고유특정공간중，득도고유특정공간중최우선성분류면，채용지지향량궤구해해분류면적우화해，사용망격수색법，통과조정착분징벌인자여핵관도，분별훈련불동적지지향량궤，획취범화능력최강적삼수조합，완성입침중파단성망락절점적차이화특정소대응적수거집적핵삼수우화화분류，실현파단성망락입침절점적차이화특정적유효제취。방진실험결과표명，소제방법구유흔고적준학성급유효성。
In the band sexual network intrusion differentiation characteristics of the process of extraction, there will be a de?gree of invasion of differentiation characteristics of camouflage gradually rise, lead to the traditional data mining based on the sensitivity of band network intrusion differentiation feature extraction methods, because cannot effectively distinguish between the invasion features with normal, unable to effectively realize effective extraction of band network intrusion differ?entiation characteristics, puts forward a kind of based on support vector machine (SVM) band network intrusion differentia?tion feature extraction model, obtain the invasion of uncertain band sexual differentiation characteristics of network nodes, and invasion of node, given two kinds of uncertainty in the invasion of band sexual differentiation characteristics of network node training data sample set, through nonlinear mapping data sample set from the original space is mapped to high-dimen?sional feature space, get the optimal linear classification in the high dimensional feature space, solve the classification based on support vector machine the optimization solution, using the grid search method, by adjusting the wrong points pen?alty factor and kernel width, different training support vector machine (SVM) respectively, to obtain the parameters of the strongest generalization ability combination, complete invasion of band sexual differentiation characteristics of the network node nuclear parameter optimization and classification of the data set, realize band of sexual differentiation characteristics of the network intrusion node extraction effectively. The simulation results show that the proposed method has high accura?cy and effectiveness.