基于改进自组织特征映射的网络入侵检测  

Network Intrusion Detection Based on the Improved SOM Algorithm

作　　者：钱景辉[1] 成伟[1] 李荣雨[1] 

机构地区：[1]南京工业大学电子与信息工程学院,江苏南京211816

出　　处：《自动化仪表》2015年第10期68-71,共4页Process Automation Instrumentation

基　　金：江苏省高校自然科学基金资助项目(编号:12KJB510007)

摘　　要：针对如何提高网络入侵检测率并进行正确分类的问题,提出了一种改进的自组织特征映射(SOM)网络算法。该算法通过对竞争机制的自适应调整来减少过度学习,并采用灰关系分析的动态权值机制降低邻域神经元中杂质的影响。KDDCUP99数据集的试验结果表明该方法具有更高的准确率。For enhancing network intrusion detection rate and implementing correct classification,the improved self-organizing map( SOM)algorithm is proposed. With this algorithm,excessive learning is decreased through adaptive adjustment of competitive mechanism,and the influence of impurities in neighborhood neurons is reduced by dynamic weights mechanism of grey relation analysis. The result of test based on KDDCUP99 data set shows that this method features higher accuracy.

关 键 词：自组织特征映射 神经网络 自适应竞争机制 灰关系 入侵检测 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]