Abstract: This paper designs exception monitoring model for Linux system. The Warn Notify(WN) protocol is designed for the model. When an exception occurs, the adoption of the protocol can be alarm information to be passed to other multiple hosts. Model uses Principal Component Analysis(PCA) combined with Hotelling’s statistics and Q statistic for anomaly detection. After an exception occurs, the contribution plots getting by principal component method is used to target a specific detection point. Experimental results show that the model has high abnormal monitoring accuracy.

Key words: exception monitoring, Principal Component Analysis(PCA), alarm transmission, contribution plots

摘要： 设计Linux系统异常监控模型。根据WN协议，当异常发生时，通过该协议将告警信息向其他多个主机传递。采用主元分析，并结合霍特林统计量及Q统计量对异常进行检测。在异常发生后，利用主元贡献图定位异常点。实验结果表明，该模型具有较高的异常监控准确率。

关键词: 异常监控, 主元分析, 告警传递, 贡献图

CLC Number: 

• N945.12