作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2013, Vol. 39 ›› Issue (1): 131-135. doi: 10.3969/j.issn.1000-3428.2013.01.028

• 网络与通信 • 上一篇    下一篇

包抽样对异常检测的影响及其消除方法研究

杨永强 1,邵 超 1,张建辉 2   

  1. (1. 河南财经政法大学计算机与信息工程学院,郑州 450002;2. 解放军信息工程大学信息技术研究所,郑州 450002)
  • 收稿日期:2012-02-27 修回日期:2012-05-15 出版日期:2013-01-15 发布日期:2013-01-13
  • 作者简介:杨永强(1974-),男,讲师、硕士,主研方向:异常流量检测;邵 超,副教授、博士;张建辉,讲师、博士
  • 基金资助:
    国家“973”计划基金资助项目(2012CB31900)

Research on Impact of Packet Sampling on Anomaly Detection and Its Elimination Method

YANG Yong-qiang 1, SHAO Chao 1, ZHANG Jian-hui 2   

  1. (1. College of Computer & Information Engineering, Henan University of Economics and Law, Zhengzhou 450002, China; 2. Institute of Information Technology, PLA Information Engineering University, Zhengzhou 450002, China)
  • Received:2012-02-27 Revised:2012-05-15 Online:2013-01-15 Published:2013-01-13

摘要: 针对当前流统计抽样在异常检测中存在的问题,分析包抽样和时域聚合导致流记录时间序列失真的原因,利用网络异常的尺度聚集特性,提出一种基于不重叠窗口聚合的多尺度分解方法,以消除由包抽样和时域聚合带来的信号噪声。仿真结果表明,该方法能降低抽样率对流尺度上信噪比的影响,提高异常检测的性能。

关键词: 异常检测, 包抽样, 时间聚合, 多尺度方法, 不重叠窗口聚合, 尺度聚集特性

Abstract: Aiming at the problem of existing flow statistical sampling in anomaly detection, this paper analyzes the distortion cause that packet sampling and time domain polymerization lead to flow record time series in theory. It uses scale accumulation feature of network anomaly, proposes multi-scale decomposition method based on Non-overlapping Window Aggregation(NOWA), eliminates the signal noise by sampling and time aggregation. Simulation result shows that this method can reduce impact of sampling rate on signal to noise ratio, improve the performance of the anomaly detection.

Key words: anomaly detection, packet sampling, temporal aggregation, multi-scale method, Non-overlapping Window Aggregation(NOWA), scale accumulation feature

中图分类号: