An analysis of graph neural networks for fake review detection: A systematic literature review

Over the past decade, detecting fake reviews has emerged as a critical challenge in ensuring the credibility of online businesses. The capability of Graph Neural Networks (GNNs) to handle data structured in graphs is a key advantage, enabling a nuanced representation of the complex relationships among reviews, users, and products. With the growing demand for robust solutions, there has been a rising focus in the research community on integrating GNNs into fake review detection mechanisms. This research conducts a thorough systematic literature review covering GNN-based articles, conference papers, and scholarly works from 2019 to January 2024, examining the current state and challenges. Initially, we provide an introduction to the problem's background and a broad overview of spam review methods based on GNNs. Additionally, we highlight existing techniques for detecting fake reviews, specifically emphasizing GNNs-based models categorized by types. Subsequently, we compare key concepts, merits, and drawbacks of these methods within their respective categories. Finally, we outline several limitations and unresolved issues that merit future attention in this domain and present potential avenues for further research. We hope this study can serve as a valuable resource for researchers, system practitioners, and newcomers to overcome current challenges and navigate forthcoming scenarios when deploying a spam review detection system with GNNs.