核聚类算法在入侵检测中的应用

APPLYING KERNEL-BASED CLUSTERING ALGORITHM IN INTRUSION DETECTION

提出了一种基于核的聚类算法,并将其应用到入侵检测中,构造了一种新的检测模型。通过利用Mercer核,我们把输入空间的样本映射到高维特征空间后,在特征空间中进行聚类。由于经过了核函数的映射,使原来没有显现的特征凸显出来,从而能够更好地聚类。而且在初始化聚类中心的选择上利用了数据分段的方法,该聚类方法在性能上比经典的聚类算法有较大的改进,具有更快的收敛速度以及更为准确的聚类。仿真试验的结果证实了该方法的可行性和有效性。

In this artic l e we proposed a kind of clustering algorithm based on kernel, and applied it in intrusion detection,thus constructed a kind of new detection model. By using Mercer kernel function,we can map the data sample in input space to a high-dimensional feature space in which we can perform clustering efficiently. Because of mapping with kernel function,those eharacters unclear before become salient, so the clustering runs better. Furthermore,we used data partition method in choosing the initialization eluster centre. This clustering method has a big improvement in performance compared with the classical clustering algorithms, and has quicker eonvergence rate as well as clustering more accurate. The results of simulation experiments show the feasibility and effectiveness of the kernel-based clustering algorithm.