网络信息审计系统中的实时网页分类方法

Categorization Method of Real-Time Web Page in Network Information Audit Systems

为了实现网络信息审计系统中的实时网页分类,提出了一种基于Dempster-Shafer证据理论的分类新方法.其基本思路是:不进行IP分片包重组,直接将网页地址特征和分片数据包作为分类的证据,计算各个证据对类的支持度,然后利用Dempster证据组合规则将各种证据提供的信息进行不断地在线融合判决,并最终给出网页的分类结果.当已有证据可以对网页进行有效分类时,对后续数据包不再做进一步处理.实验结果表明,所提方法的查准率大于83%,查全率大于90%,在分类性能和运行时间上均优于已有的基于分片的模糊K最近邻分类算法.

To accomplish the categorization of real-time web page in network information audit systems, a new method based on Dempster-Shafer evidence theory is proposed. The main idea of it is as follows： the web page address and fragments are regarded as the evidence of categorization without reassembling IP fragments, and then the support degree that each of evidence stands up for a category is calculated. The Dempster combination rule is used to continuously fuse and adjudge the information provided by various evidences online, and finally the categorized result is obtained. When the existing evidences can efficiently categorize the web page, the subsequence fragments are no need to be handled further. The experiment shows that the precision rate and recall rate of the proposed method are larger than 83% and 90% respectively, and it is superior to the fuzzy K nearest neighbor algorithm based on fragments in the categorization performance and running time.