有监督SOM神经网络在入侵检测中的应用

摘要

In order to improve the classification effectiveness of Self-organizing feature Mapping(SOM) neural network, this paper designs a Supervised Self-organizing feature Mapping(SSOM) network, which adds output layer into the network structure based on input layer and competitive layer. According to different prediction category of input samples, SSOM selects different formula to adjust weights and train network. Through combinations of two weights, SSOM realizes the regression and statistic of the sample classes. Experiments on KDD CUP99 dataset show that SSOM has better classification ability and higher intrusion detection rate.%为提高自组织特征映射(SOM)神经网络的分类性能,提出一种有监督SOM神经网络(SSOM).在输入层和竞争层的基础上增加输出层,根据输入样本的不同预测类别,选取不同的公式调整权值,并训练网络.通过2个权值的组合,实现对样本类别的回归和统计.基于KDD CUP99入侵检测数据集的实验结果表明,与其他SOM网络相比,SSOM具有更好的分类性能和更高的入侵检测率.