Robust features for trustable aggregation of online ratings

Abstract

When purchasing an online product, customers tend to be influenced strongly by its reputation, the aggregation of customers' ratings on the product. The reputation, however, is not always trustable since it can be easily manipulated by attackers. In this paper, we first address identifying trustable users on a given product in online rating systems, and computing its true reputation by aggregating only their ratings. In order to find these trustable users, we list candidate user features significantly related to the trustworthiness of users and verify the robustness of each user feature through extensive experiments.