网络与信息安全学报 ›› 2015, Vol. 1 ›› Issue (1): 66-71.doi: 10.11959/j.issn.2096-109x.2015.00009

• 学术论文 • 上一篇    下一篇

基于改进聚类分析的网络流量异常检测方法

李洪成1(),吴晓平1,姜洪海2   

  1. 1 海军工程大学 信息安全系,湖北 武汉 430033
    2 海军北海舰队 司令部,山东 青岛 266071
  • 修回日期:2015-10-08 出版日期:2015-12-01 发布日期:2016-01-12
  • 作者简介:李洪成(1991-),男,河南商丘人,海军工程大学博士生,主要研究方向为信息安全、数据挖掘。|吴晓平(1961-),男,山西新绛人,博士,海军工程大学教授、博士生导师,主要研究方向为信息安全、密码学。|姜洪海(1972-),男,山东乳山人,海军北海舰队司令部工程师,主要研究方向为信息安全。
  • 基金资助:
    国家自然科学基金资助项目(61100042);中国博士后基金资助项目(2014M552656);湖北省自然科学基金资助项目(2015CFC867)

Traffic anomaly detection method in networks based on improved clustering algorithm

Hong-cheng LI1(),Xiao-ping WU1,Hong-hai JIANG2   

  1. 1 Information Security Department,Naval University of Engineering,Wuhan 430033,China
    2 Headquarters,Command of Naval North-Sea Fleet,Qingdao 266071,China
  • Revised:2015-10-08 Online:2015-12-01 Published:2016-01-12
  • Supported by:
    The National Natural Science Foundation of China(61100042);Postdoctoral Science Foundation of China(2014M552656);The Natural Science Foundation of Hubei Province(2015CFC867)

摘要:

针对传统基于聚类分析的网络流量异常检测方法准确性较低的问题,提出了一种基于改进 k-means聚类的流量异常检测方法。通过对各类流量特征数据的预处理,使k-means算法能适用于枚举型数据检测,进而给出一种基于数值分布分析法的高维数据特征筛选方法,有效解决了维数过高导致的距离失效问题,并运用二分法优化K个聚簇的划分,减少了初始聚类中心选择对k-means算法结果的影响,进一步提高了算法的检测率。最后通过仿真实验验证了所提出算法的有效性。

关键词: 网络安全, 流量异常检测, 聚类分析, k-均值算法

Abstract:

To solve the problem that traditional traffic abnormal detection methods were not accurate enough,a traf-fic anomaly detection method based on improved k-means was proposed.All kinds of network traffic data were pre-processed to make k-means algorithm can apply to enumeration data detection.Then a features selection method was pro-posed with the analysis of the distribution of network traffic data to avoid the distance useless caused by too much fea-tures.Furthermore,the clustering process of K clusters was optimized based on dichotomy,aiming to reduce the effects of initial clusters centers selection.Simulation results demonstrate the effectiveness of the algorithm.

Key words: network security, traffic abnormal detection, clustering analysis, k-means algorithm

中图分类号: 

No Suggested Reading articles found!