STA : Sybil Types-aware Robust Recommender System

Abstract

Recently many users refer to various recommender sites when they buy things, movies, music and etc with a rapid development of internet. But there are malicious users (Sybil) to raise or lower ratings of items intentionally in these recommender sites, finally recommender system can recommend incomplete or inaccurate results to normal users. We suggest a recommender algorithm to separate ratings which users generate into normal ratings and outlier ratings and to minimize effects of malicious users. In addition, it provides stable RS about three kinds of models (Random attack, Average attack and Bandwagon attack) which are making problems in Recommender system now. To prove performances of suggesting method, we conducted performance analysis to collect real data (crawling). As a result of performance analysis, it is proved that a performance of suggesting method is good regardless of Sybil size compared to existing algorithms.