Fake Review Detection via Exploitation of Spam Indicators and Reviewer Behavior Characteristics

Abstract

The rapid spread of Internet technologies has redefined E-commerce, since opinion sharing by product reviews is an inseparable part of online purchasing. However, e-commerce openness has attracted malicious behaviors often expressed by fake reviews targeting public opinion manipulation. To address this phenomenon, several approaches have been introduced to detect spam reviews and spammer activity. In this paper, we propose an approach which integrates content and usage information to detect fake product reviews. The proposed model exploits both product reviews and reviewers’ behavioral traits interlinked by specific spam indicators. In our proposed method, a fine-grained burst pattern detection is employed to better examine reviews generated over “suspicious” time intervals. Reviewer’s past reviewing history is also exploited to determine the reviewer’s overall “authorship” reputation as an indicator of their recent reviews’ authenticity level. The proposed approach is validated with a real-world Amazon review dataset. Experimentation results show that our method successfully detects spam reviews thanks to the complementary nature of the employed techniques and indicators.