Mitigating Harm of Liar-Farm in Reputation Model of VoIP Spam Filtering System

Abstract

Reputation can help VoIP system detect spammers and filter spam calls, however, existing of liar farm would destroy the infrastructure of reputation system. As the spammers are more and more trickish, they would certainly construct liar farms and ask liars to inject unfair positive evaluation to raise their reputation. Furthermore, false recommendation emerged dynamically and randomly. Hence, most traditional solution can hardly mitigate the threat from liars. In this paper, we proposed a novel schema which is based on game theory against false recommendation. In our approach, we define recommendation game (RG) and construct the strategy of reputation requester, "tit for tat", against dynamical and random behavior of spammer and liar. At last, we verify the RG in the P2P-AVS (anti-voice-spam). Results show that RG could mitigate the threat of liar farm and raise the accuracy of spam detection, and make reputation system be robust and stable even if there are a lot of liars.