一种抽取多数类边界样本的入侵检测算法  

An Intrusion Detection Algorithm by Selecting Bound Vectors of Dominated Class

作　　者：彭胜伟[1] 

机构地区：[1]河南化工职业学院,郑州450042

出　　处：《科技通报》2013年第10期106-108,共3页Bulletin of Science and Technology

摘　　要：针对传统网络入侵检测方法在实时性响应和入侵行为识别率上存在的不足,本文提出了一种抽取多数类边界样本的入侵检测算法。该算法首先根据中心距离确定网络链接行为中多数类样本的边界样本,然后将多数类样本的边界样本与少数类样本合并构成新的训练集合,最后进行分类学习。该算法有效地降低了类别之间的不平衡度和减少了训练样本数目,具有更好的入侵检测性能。在KDD CUP99数据集上的仿真实验,充分验证了该算法的有效性。As traditional network intrusion detection methods are not good at real-time response and intrusion recognition rate, this paper proposes an intrusion detection algorithm by selecting bound vectors of dominated class. Firstly, determine the bound vectors of dominated class according to the center distance. Then combine the bound vectors of dominated class with the minority class to construct a new training set. Finally, do classification on the new set. As the imbalance degree and the number of training samples are effectively reduced, the algorithm has a better performance of intrusion detection. The simulation experiments on KDD CUP 99 data set fully verify the effectiveness of the algorithm.

关 键 词：入侵检测 实时性 边界样本 中心距离 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]