摘要
针对目前入侵检测系统误报率过高、检测率不高和对未知入侵检测能力有限的缺陷,提出一种动态SOFM的网络入侵检测方法,定义了聚类节点信任度,并根据竞争结果、信任度、中心相似度,制定节点的增删策略,提升聚类效果。使用KDD99数据集进行实验,结果表明系统在保持误报率低的情况下,入侵检测率有所提高。
Considering current intrusion detection system with high misinformation rate and low detection rate, this paper applies dynamic SelfOrganizing Feature MAp（SOFM）neural network to intrusion detection , it has defined clustering node trust degree , owing to competing result trust degree, the centre similarity degree, it can work out the node addition and deletion trActics, raise clustering result.we use the KDD99 data collection to carry out An experiment,the result indicates systematically under keeping low condition of misuse rate , intrusion detecting rates improve to some extent.
出处
《计算机安全》
2009年第8期22-24,共3页
Network & Computer Security
关键词
入侵检测
神经网络
自组织特征映射
intrusion detection
Neural networks
self-organizing feature maps
