提出一种新颖的基于服务特征分析的入侵检测方法。在处理网络审计数据时,首先针对网络服务进行特征分析,将审计数据按照网络应用进行区分,然后使用统计方差模型对应用区分后的审计数据进行检测;另外,在传统的统计方差模型基础上,提出加权的方法调整可信区间,提高检测率。选用KDDCup 1999 Data网络连接数据集进行实验,基于服务特征分析与统计的入侵检测方法在不增加虚警率的情况下,可以得到更高的检测精度。结果说明,该方法是行之有效的。
A new method for intrusion detection is proposed baaed on network service characteristics. While processing the network audit data,do the characteristics analysis with network services and differentiate the audit data by network applications. Then finish the detection with statistics variance model. What's more, while adopting the traditional statistics variance model change the credibility interval by weight to improve the performance. At last, use the data set of KDDCup' 99, and the result of experiment shows the proposed method could achieve better detecting rates.