What Is the Internet Water Army? A Practical Feature-Based Detection of Large-Scale Fake Reviews

Abstract

Information is passed by word-of-mouth figures prominently when consumers evaluate products through reviews. However, severe logistical problems are caused by the internet’s Water Army (i.e., literally people who are hired by individuals or organizations to compose false reviews), that flood the internet e-commerce websites. An array of internet e-commerce sites is flooded with inauthentic information, and false reviews are used to maliciously induce consumers to purchase specific products, that often contain some defects. Notwithstanding the fact that the internet Water Army first manifested in China, it can also exist in other countries. The rationale lies in the high profitability possible, in the minds of numerous organized underground paid poster groups, and in writing fake reviews to misinform consumers. It has become an increasingly daunting task to precisely spot the Water Army members, who often alter their writing style and posted content. In this paper, the authors devise a comprehensive set of features to characterize all users and compare the paid posters against the normal users on different dimensions; furthermore, an ensemble detection model equipped with seven disparate algorithms is put into place. Our model reached a score of 0.730 in the AUC measure, 0.691 in the F1 measure on the JD dataset, 0.926 in the AUC measure, and 0.871 in the F1 measure on the Amazon dataset, which outshines the measures in the existing research. The significance and contribution of this work are in advancing constructive solutions and recommendations for this major concern of the entire e-commerce industry.