基于遗传神经网络的入侵特征选择模型

An Intrusion Feature Selection Model Based on Genetic Neural Network

针对高维数据包含的不相关和冗余特征影响检测方法性能的问题,提出了基于遗传神经网络入侵特征选择模型。该模型在传统的遗传神经网的输入层和隐含层之间增加特征选择层,并在特征选择层与输入层间设置连接开关,如果开关合上,则该特征被选中;否则为放弃。实验结果表明,该模型在保持原有信息完整性的同时,能有效减少冗余特性;在保证检测准确率的前提下,有效提高系统的检测速度。

The performance of a detection method was degraded by the high dimensional data containing irrelevant and redundant attributes. This paper proposed an intrusion feature selection model based on genetic neural network. The feature selection layer was inserted in between input layer and hidden layer of the traditional genetic neural network, and the connection switch was set between the input layer and feature selection layer. If the switch state is off, the features are selected; otherwise the features are given up. Experimental results showed that this approach can main- tain the original integrity of the information, effectively reduce the redundancy. It can improve the detection speed effectively with high detection accuracy.