基于IP地址聚类的反垃圾邮件信誉系统  

IP clustering based reputation system for anti-spam

作　　者：张洪[1] 段海新[1] 吴建平[1] 

机构地区：[1]清华大学计算机科学与技术系,北京100084

出　　处：《清华大学学报（自然科学版）》2010年第10期1723-1727,共5页Journal of Tsinghua University(Science and Technology)

基　　金：国家"九七三"基础研究项目(2009CB320505);国家信息安全项目((242)2007A07)

摘　　要：现有基于用户信誉的反垃圾邮件系统存在诸如发送方标识容易被篡改等缺点,使得恶意用户可以通过特定行为逃避信誉系统的管制。针对这一现象,提出了一种基于IP地址聚类的反垃圾邮件信誉系统(RepuRank)。该系统根据电子邮件的源IP地址进行聚类,并对这些类进行信誉评价,然后利用得到的信誉值来判断新收到的电子邮件是否为垃圾邮件。实验采用了公认的反垃圾邮件指标对RepuRank的性能进行评估,结果表明RepuRank不仅能有效地区分正常邮件与垃圾邮件,还能防止恶意用户使用特定行为逃避信誉系统管制的问题。The existing reputation systems for anti-spam have some limitations that make them vulnerable to be evaded.This paper presents a reputation system named RepuRank to deal with this problem.Instead of computing the reputation value for a single sender,RepuRank clusters the senders into groups based on their IP addresses,computes the reputation values of the obtained groups,and classifies the following messages with the reputation values.Based on the publicly available anti-spam evaluation metrics,experimental results show that RepuRank can not only effectively separate legitimate messages from spam messages,but also prevent the proposed system from being evaded by malicious users.

关 键 词：网络安全 IP地址聚类 反垃圾邮件 信誉系统 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]