一种基于改进ART-2的入侵检测方法  

A New Method in IDS Based on ART-2 Neural Network

作　　者：刘军利[1] 陈莉[1] 

机构地区：[1]西北大学计算机科学系,陕西西安710069

出　　处：《微机发展》2005年第4期119-121,128,共4页Microcomputer Development

基　　金：陕西省自然科学基金资助项目(98X11);陕西省教育厅重点科研计划项目(00JK015)

摘　　要：指出了直接把标准ART-2网络应用于入侵检测时存在的两个问题:对基本相似、仅有个别分量差别较大的向量不能正确分类;输入向量特征丢失。并根据入侵检测的特定应用,相应地提出在首先对输入向量进行规范化处理,然后用新引入的一种具有更严格测试准则ART-2网络对其进行处理的方法,以期提高入侵检测系统的检测率和误检率。In this paper, by analyzing the structure of standard ART-2 network, two problems caused by using it directly in IDS are indicated. The first problem is standard ART-2 network unable to distinguish two vectors when they are very resemble in spite of several is dissimilar greatly. Secondly, after processed by F1 layer, the character of vector may be lost. Considering the characteristic of IDS, we propose that the vectors need to be standardized beforehand and then introduce a new ART-2 network which has more vigorous vigilance test criterion to process them, in order to improve the performance of IDS.

关 键 词：入侵检测系统 神经网络 ART-2 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]