多神经网络组合模型在入侵检测中的应用

Intrusion Detection Technique Based on a Combination Model of Multi Neural Networks

入侵检测技术作为提高网络安全的有效手段日益受到重视,利用多神经网络组合模型解决入侵检测问题,针对入侵检测研究的通用审计数据集,首先将其所有字符串行式的元素转换为数值形式;其次为了提高神经网络的逼近性能和运算速度,去除对输出无影响的输入项,并且将剩余输入项的可能取值转换到合理的范围内;最后在MATLAB平台下进行仿真实验,并与单层BP网络进行比较.仿真结果表明,多神经网络组合模型在入侵检测中体现出良好的特性.

Intrusion detection technique is an important way to improve network security, and is a main research topic in computer field. In this paper, the benchmark dataset commonly used in the research of intrusion detection is adopted. Firstly, the string type of data is changed into the numeric type ; secondly the unnecessary data is omitted and the value of data is limited in reasonable ranges, which can improve the performance of the network ; finally a lot of simulation results on a combination model of RBF and BP networks are gained from Matlab platform, and comparison results display that multi neUral network is better than BP network for intrusion detection.