IP clustering based reputation system for anti-spam

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.