计算机工程 ›› 2010, Vol. 36 ›› Issue (10): 158-159.doi: 10.3969/j.issn.1000-3428.2010.10.053

• 安全技术 • 上一篇    下一篇

最大似然估计ROC凸包曲线算法

梁海华,李克清,盘丽娜   

  1. (常熟理工学院计算机科学与工程学院,常熟 215500)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2010-05-20 发布日期:2010-05-20

Maximum Likelihood Estimation of ROC Convex Hull Algorithm

LIANG Hai-hua, LI Ke-qing, PAN Li-na   

  1. (College of Computer Science and Engineering, Changshu Institute of Technology, Changshu 215500)
  • Received:1900-01-01 Revised:1900-01-01 Online:2010-05-20 Published:2010-05-20

摘要: 采用列联表表示特定阈值下入侵检测系统(IDS)的性能,使用ROC曲线对不同阈值下IDS的总体性能进行评估。在实际应用中,基于样本数据集,通过计算几何的方式得到ROC凸包曲线(ROCCH)。为降低ROCCH的计算复杂度,在曲线下面积(AUC)最大似然估计的条件下,通过保序回归得到最大似然估计ROC曲线(MLE-ROC)算法。实验表明,MLE-ROC算法在降低计算复杂度的同时提高了AUC的近似程度。

关键词: 入侵检测系统, 列联表, 最大似然估计ROC凸包曲线, 保序回归

Abstract: The measures of Intrusion Detection System(IDS) in special threshold-value are derived from the contingency table. ROC curve facilitates to evaluate general performance of IDS. In practice, ROC Convex Hull(ROCCH) method which combines techniques from computational geometry is used. In order to low computational complexity, the algorithm of MLE-ROC which depending on monotonic regression under maximum likelihood estimation of value of Area-Under-Curve(AUC) is proposed. Experimental results show that the proposed algorithm is computationally efficient while having better AUC approximation level.

Key words: Intrusion Detection System(IDS), contingency table, MLE-ROCCH, monotonic regression

中图分类号: