Graph-based review spammer group detection

摘要: Online product reviews nowadays are increasingly prevalent in E-commerce websites. People often refer to evaluate the quality of a before purchasing. However, there have been large number review spammers who work collaboratively promote or demote target products, which severely harm system. Much previous exploits machine learning approaches detect suspicious reviews/reviewers. In this paper, we introduce top-down computing framework, namely GGSpam, spammer groups by exploiting topological structure underlying reviewer graph reveals co-review collusiveness. A novel instantiation GSBC, is designed modeling as bi-connected graphs. Given graph, GSBC identifies all components whose spamicity scores exceed given spam threshold. For unsuspicious graphs, minimum cut algorithm used split and smaller graphs further processed recursively. variety group indicators measure group. Experimental study shows that proposed approach both effective efficient outperforms several state-of-the-art baselines, including based non-graph based, margin.