Modeling and Evaluating a Robust Feedback-Based Reputation System for E-Commerce Platforms

Abstract

Despite the steady growth of e-commerce communities in the past two decades, little has changed in the way these communities manage reputation for building trust and for protecting their member's financial interests against fraud. As these communities mature and the defects of their reputation systems are revealed, further potential for deception against their members is created, that pushes the need for novel reputation mechanisms. Although a high volume of research works has explored the concepts of reputation and trust in e-communities, most of the proposed reputation systems target decentralized e-communities, focusing on issues related with the decentralized reputation management; they have not thus been integrated in e-commerce platforms. This work's objective is to provide an attackresilient feedback-based reputation system for modern e-commerce platforms, while minimizing the incurred financial burden of potent security schemes. Initially, we discuss a series of attacks and issues in reputation systems and study the different approaches of these problems from related works, while also considering the structural properties, defense mechanisms and policies of existing platforms. Then we present our proposition for a robust reputation system which consists of a novel reputation metric and attack prevention mechanisms. Finally, we describe the simulation framework and tool that we have implemented for thoroughly testing and evaluating the metric's resilience against attacks and present the evaluation experiments and their results. We consider the presented simulation framework as the second contribution of our article, aiming at facilitating the simulation and elaborate evaluation of reputation systems which specifically target e-commerce platforms by thoroughly presenting it, exhibiting its usage and making it available to the research community.