信息网络安全 ›› 2019, Vol. 19 ›› Issue (9): 101-105.doi: 10.3969/j.issn.1671-1122.2019.09.021

• • 上一篇    下一篇

基于前馈神经网络的入侵检测模型

冯文英1,2, 郭晓博2, 何原野2, 薛聪2   

  1. 1.中国科学院大学网络空间安全学院,北京 100049
    2.中国科学院信息工程研究所,北京 100093
  • 收稿日期:2019-07-15 出版日期:2019-09-10 发布日期:2020-05-11
  • 作者简介:

    作者简介:冯文英(1995—),女,河北,博士研究生,主要研究方向为信息安全、深度学习;郭晓博(1990—),女,河北,工程师,硕士研究生,主要研究方向为信息安全;何原野(1995—),男,河南,研究实习员,硕士研究生,主要研究方向为数据挖掘;薛聪(1990—),女,河北,助理研究员,博士研究生,主要研究方向为事件数据挖掘。

  • 基金资助:
    国家自然科学基金[U163620068]

Intrusion Detection Model Based on Feedforward Neural Network

Wenying FENG1,2, Xiaobo GUO2, Yuanye HE2, Cong XUE2   

  1. 1. School of Cyber Security, University of Chinese Academy of Sciences, Beijing 100049, China
    2. Institute of Information Engineering, Chinese Academy of Sciences, Beijing 100093, China
  • Received:2019-07-15 Online:2019-09-10 Published:2020-05-11

摘要:

由于入侵行为特征多样、网络环境复杂,导致基于深度学习的入侵检测方法容易出现模型复杂、灵活性差等问题。为此,文章提出基于前馈神经网络的入侵检测模型SFID,通过逐层削减神经元数量,整体化解决特征抽取和入侵分类问题,从而降低了入侵检测模型的训练复杂度。通过实验验证,模型在正确率相当的情况下比S-NDAE模型训练效率明显提高。

关键词: 入侵检测, 前馈神经网络, 误差反向传播算法

Abstract:

However, due to the diversity of intrusion behavior features and the complex network environment, intrusion detection methods based on deep learning are prone to have complex models and poor flexibility. To solve this problem, this paper proposed an intrusion detection model called SFID (Simplified Feedforward Intrusion Detection) based on feedforward neural network, which can integrate feature extraction and intrusion classification by reducing the number of neurons layer by layer, thus simplify the training complexity of intrusion detection model. With the verification, the training efficiency of this model is higher than that of S-NDAE model under the same accuracy.

Key words: intrusion detection, feedforward neural network, error backpropagation algorithm

中图分类号: