本文已被:浏览 1437次 下载 479次
投稿时间:2013-09-30
投稿时间:2013-09-30
中文摘要: 针对传统BP神经网络在检测速度、精度、复杂度等方面的缺陷,提出了一种基于深度信念网(deep beliefnets,DBN)的网络入侵检测算法,将数据通过双层RBM结构降维,再用BP神经网络反向微调结构参数,从而简化了数据复杂度,减少了BP神经网络的计算量.通过对KDD99数据集仿真实验表明,该算法对于大数据拟合快,检测精度较高.
Abstract:DBN(deep belief nets)-based intrusion detection algorithm is proposed for improving the traditional BP neural network in detection speed,accuracy,complexity,etc.The BP neural network structure will do a fine turning on parameters after the double RBMs structure reducing the dimension of data.In this way,the complexity of data and BP neural network is simplified.The experiments with KDD99 dataset show that the new algorithm is in excellent performance in large data fitting and intrusion detection.
文章编号:20130617 中图分类号: 文献标志码:
基金项目:信息安全国家重点实验室(中国科学院软件研究所)开放式基金(04-02-1);上海市教育委员会创新基金(11YZ192);上海市“科技创新行动计划”重点项目(11511504400)
引用文本:
徐东辉,王勇,樊汝森.一种基于DBN的网络入侵检测算法[J].上海电力大学学报,2013,29(6):589-592.
XU Donghui,WANG Yong,FAN Rusen.A Network Intrusion Detection Algorithm Based on DBN[J].Journal of Shanghai University of Electric Power,2013,29(6):589-592.
徐东辉,王勇,樊汝森.一种基于DBN的网络入侵检测算法[J].上海电力大学学报,2013,29(6):589-592.
XU Donghui,WANG Yong,FAN Rusen.A Network Intrusion Detection Algorithm Based on DBN[J].Journal of Shanghai University of Electric Power,2013,29(6):589-592.