基于模糊SVM模型的入侵检测分类算法

为解决入侵检测分类遇到的训练样本数量少、分类准确率低的问题，提出基于模糊支持向量机的多级分类机制。该分类机制训练模糊SVM模型将数据粗分为正常与攻击大类，采用DBSCAN算法产生细分模型进行攻击子集的自动聚类，将有关数据细分得到攻击的具体细类。在机制设计中，优化了隶属度函数的计算、设计了数据标准化与归一化等过程，并训练了高效分类器。实验表明，针对网络入侵检测数据中常见的孤立点干扰、噪声多，并且负样本占比多的网络业务数据集，新算法在保持分类准确率高的前提下，分类过程的计算时间较短。

IDS classification algorithm based on fuzzy SVM model

In order to solve the problem of small number of training samples and low classification accuracy in intrusion detection classification, this paper proposed a multi-level classification mechanism based on fuzzy support vector machines(FSVM). This classification mechanism firstly trained the fuzzy SVM model to divide the data roughly into normal and attack categories, and then used DBSCAN algorithm to generate subdivision model for automatic clustering of attack subsets, and then subdivided the relevant data into specific classes of attack. In the mechanism design, it optimized the calculation of membership function, and designed the process of data standardization and normalization, and trained the efficient classifier. Aiming at the network service data sets with frequent isolated point interference, much noise and a large proportion of negative samples in the network intrusion detection datasets, experiments show that the new algorithm requires less computing time in the process of classification under the premise of high classification accuracy.