Generating unfair ratings to attack online rating systems II

Abstract

Online rating systems provide users with information to make decisions. However,

these systems may be influenced by malicious behaviours from buyers and sellers.

Attack models are proposed to address various scenarios based on how dishonest

buyers give unfair ratings to sellers. Trust models are proposed by researchers to

analyze and filter these unfair ratings. In this project, A testbed is programmed to

evaluate effectiveness of trust models against attack models with both simulated

environment data and real date collected from online marketplace. Single criteria

rating system and multi-criteria rating system are considered in this project. The

experiment results provide a comprehensive comparison of the performance of trust

models under different attacking scenarios.