基于聚类集成的网络入侵检测算法

摘要

为了进一步提高网络入侵检测的效果,提出一种基于聚类集成的入侵检测算法.首先利用Bagging算法从训练集中生成多个训练子集.然后调用模糊C均值聚类算法训练并生产多个基本聚类器.然后利用信息论构造适应度函数.采用粒子群算法从上述聚类集体中获得一个具有最优性能的集成聚类器.仿真实验结果表明,该算法能有效的提高入侵检测的精度,具有较高的泛化性和和稳定性.%To improve the ability of network intrusion detection, a detection algorithm based on clustering ensemble is presented. First, many training subsets were produced from training dataset by Bagging, and clustering individuals were trained by fuzzy c-means clustering. Then, fitness function was construct using information theory, ensemble clustering machine of better ability were obtained from clustering individuals based on particle swarm optimization algorithm. The experiments show that the algorithm effectively improve accuracy of intrusion detection, it have higher generalization performance and stability.