带野值的单类分类器在安全审计中的应用

Application of one-class classifier with negatives in security audit data analysis

单类分类器是当前模式识别领域的一个研究热点。带野值的单类分类器是在单类分类器的基础上,通过引入少量珍贵的异常样本(野值),以加强分类器的性能。该模型适用于处理正类样本数目远多于反类样本的两类数据类别不平衡问题。提出了将带野值的支持向量描述方法应用于安全审计数据分析中,并通过实验证实了该方法对异常样本更为敏感,具有良好的应用潜力。

One-class classifier is currently a hot spot of pattern recognition field.One-class classifier with negatives is based on one-classifier,by leading into a few costful abnormal samples to reinforce the classification.This model applies to the problems handling the two kind data categories imbalances where positives more over than negatives.It is proposed in this paper that using support vector data description with negatives in security audit data analysis system.Through some experiinents,it is proved to be more sensitive with exceptional samples,so it will be more valuable in practice.