A Weighted Raw Reputation Generating Approach Based on Similarity

Abstract

In the anti-spam field, raw reputation is the current mailing behavior of one email server. Meanwhile, it is the foundation of the distributed spam processing technology based on reputation mechanism. In this paper, the advantages and disadvantages of the existing several raw reputation generating approaches are analyzed, and a new method: MSGuard is proposed. MSGuard is a weighted raw reputation generating approach based on similarity. Simulation results demonstrate that: in the scenario which the malicious nodes provide inauthentic evaluations, the average differences between the expectations and the raw reputations calculated by TrustGuard and MSRep are 0.4 and 0.5 respectively. And the difference of either EigenTrust or MSGuard is only approximate 0.05. In the scenario which the collusive and disguised malicious nodes exist, the difference between the expectation and the raw reputation calculated by EigenTrust is 0.25, and it is less than 0.1 by MSGuard. MSGuard can reflect nodes’ actual mailing situations more accurately.