Traditional methods ignore the imbalance of network data, resulting in unsatisfactory clustering detection results, long detection time, and high rate of missed detection and false alarm. In this regard, this paper proposes a clustering detection method of network intrusion feature based on support vector machine and LCA block algorithm. Firstly, the useless features were deleted by reducing the dimension of the data set, thus improving the clustering detection accuracy. Secondly, the training sample set was divided, and the multi-level support vector model was established by two classification support vector machines. Finally, the LCA algorithm was adopted to identify the network intrusion features, achieving clustering detection of network intrusion feature. The results show that the proposed method achieves better clustering detection results and effectively reduces the average clustering detection time.

传统方法忽略了网络数据的不平衡，导致聚类检测结果不理想，检测时间长，漏检率高，误报率高。为此，本文提出了一种基于支持向量机和LCA块算法的网络入侵特征聚类检测方法。首先，通过减少数据集的维数来删除无用的特征，从而提高聚类检测的准确性。其次，将训练样本集进行划分，并通过两个分类支持向量机建立多级支持向量模型。最后，采用LCA算法对网络入侵特征进行识别，实现了对网络入侵特征的聚类检测。