Fake reviews deliberately created and propagated as part of disinformation campaigns pose a significant threat to consumers, as they can lead them astray and result in financial losses and reputational damage. It is crucial to detect and mitigate these deceptive practices to maintain trust and integrity on online platforms. In this study, we propose a novel approach to enhance the effectiveness of fake review detection using GraphSAGE(Graph Sample and Aggregate), a graphbased technique, with a Covertness model. The innovative integration of GraphSAGE with a covertness measure aims to capture the intricate interactions and heterogeneity present in user reviews, by considering both user attributes and textual content. Our performance evaluations on publicly available real-world Amazon datasets demonstrate that our proposed model consistently achieves competitive Recall and AUC values across various training data percentages. Additionally, our comparative analysis against other graph-based models, including GCN, GAT, GeniePath, and GraphSAGE, highlights the superior performance of our proposed model. Our findings emphasize the robustness and potential of our model to accurately detect fake reviews in real-world scenarios. This study significantly contributes to advancing fake review detection methodologies by offering a promising approach to combat disinformation and safeguard consumer trust on online platforms.

PDF