自动确定聚类数算法在网络入侵检测中的应用

发布时间：2018-08-04 21:54

【摘要】：研究准确检测入侵问题,为保护网络资源的安全,在入侵检测中需要预先指定聚类数目以及对初始中心敏感,但传统K均值算法易陷入局部最优。为解决上述问题,提出了一种自动确定聚类数的算法,首先通过多次执行样本抽样技术用最大最小距离算法产生一系列较优的聚类中心和聚类数目,作为差分进化算法的初始种群,然后在种群进化过程中以最优种群个体为指引,动态调整个体聚类中心和聚类数目,并且利用差分进化算法的全局寻优能力和K均值算法的局部搜索能力,对聚类中心和聚类数目同时进行优化,从而得到最佳的聚类划分和聚类数目。在上述算法的基础上,提出了一种入侵检测判断方法,通过对KDD CUP1999数据集的仿真结果表明,改进算法具有较好的入侵检测效果,能够有效的检测出网络中的入侵数据。
[Abstract]:In order to protect the security of network resources, it is necessary to specify the number of clusters and be sensitive to the initial center in intrusion detection, but the traditional K-means algorithm is prone to fall into local optimum. In order to solve the above problems, an algorithm for automatically determining the number of clusters is proposed. Firstly, a series of optimal clustering centers and clustering numbers are generated by using the maximum and minimum distance algorithm through the multi-execution sample sampling technique. As the initial population of the differential evolution algorithm, and guided by the optimal population individuals during the evolution of the population, the cluster center and the number of individual clusters are dynamically adjusted. The global optimization ability of differential evolution algorithm and the local search ability of K-means algorithm are used to optimize the clustering center and the number of clusters at the same time, so as to obtain the best clustering partition and clustering number. On the basis of the above algorithms, an intrusion detection judgment method is proposed. The simulation results of KDD CUP1999 dataset show that the improved algorithm has better intrusion detection effect and can effectively detect intrusion data in the network.
【作者单位】：江南大学物联网工程学院;
【分类号】：TP393.08

【参考文献】