北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2011, Vol. 34 ›› Issue (2): 45-49.doi: 10.13190/jbupt.201102.45.yuyh

• 论文 • 上一篇    下一篇

网络异常点检测中性能指标阈值的动态确定方法

于艳华,宋美娜,张文婷,宋俊德   

  1. 北京邮电大学 计算机学院, 北京 100876
  • 收稿日期:2010-04-23 修回日期:2010-12-09 出版日期:2011-04-30 发布日期:2011-04-28
  • 通讯作者: 于艳华 E-mail:yhyu_bupt@sina.com
  • 基金资助:

    国家“十一五”科技支撑项目(2008BAH24B04); 国家自然科学基金项目(61072060); 中央高校基本科研业务费专项资金项目

A Dynamic Compuation Approach to Determining the  Threshold in Network Anomaly Detection

  • Received:2010-04-23 Revised:2010-12-09 Online:2011-04-30 Published:2011-04-28
  • Supported by:

    ;the National Key Technologies Research and Development program of China during the 11th;supported by Chinese Universities Scientific Fund

摘要:

通过检测代表性能降质的异常点来实现故障的提前发现和快速恢复是提高通信网的可靠性的重要手段. 采用基于统计假设检验的网络异常点检测方法,提出一种综合运用季节累积自回归滑动平均模型时间序列预测和置信区间计算来动态获取性能指标阈值的方法. 利用累积自回归滑动平均模型在训练集上的拟合残差白噪声符合正态分布的假设,给出了一种通过构造满足t分布的随机变量来计算预测值在任意置信度1-α下置信区间的新算法. 理论分析和实验结果表明,该阈值动态确定方法有效.

关键词: 异常点检测, 时间序列预测, 季节累积自回归滑动平均模型, 置信区间

Abstract:

Fault correction by detecting anomalies designating performance degradation is an important approach to improving the reliability of communication network. Statistical hypotheses testing approach is employed to detect network anomaly. A new approach to acquiring the fluctuation threshold is proposed comprehensively when taking advantage of time series prediction confidence interval computation based on multiplicative autoregressive integrated moving average. Furthermore, under the assumption that the training residual which is a white noise process follows normal distribution, the associated confidence interval of prediction can be figured out under any given confidence degree by constructing random variable satisfying t distribution. Experiments verify the effectiveness of anomaly detection mechanism and the accuracy of the algorithm. 

Key words: anomaly detection, time series prediction, seasonal autorgressive integrated moving average, confidence interval