基于混合神经网络的入侵检测技术

本文将自组织映射神经网络与学习矢量量化的学习算法混合,用于基于程序行为的本机入侵异常检测中。考虑到目前很多方法存在成功率低以及训练时间长的缺点,本文利用自组织映射神经网络对数据聚类,然后通过学习矢量量化对已聚类的数据再进行分类。仿真实践证明,这种混合优化技术可使分类的边界得以收缩,可提高分类精度和准度,提高了入侵检验的成功率。

Intrusive Detection Technology Based on Mixed Neural Networks

This paper proposes an advanced method which was mixed by self- organizing map neural network and learning vector quan- tization algorithm,and used in intrusive detection system based on program behavior of local machine. For improving its low success possibility and long training time,this paper makes use of self- organizing map neural network to cluster data and then classifies these data by learning vector quantization.Using a simulation example,the mixed method not only cleared the edge of classification, ad- vanced precision and accuracy of classification,but also improved its success possibility of intrusive detection.