Key node propagation-based overlapping spammer group detection algorithm on e-commerce platforms

With the rapid growth of e-commerce platforms, spammer groups have increasingly used fake reviews to influence consumer decisions, posing significant challenges to platform governance. This issue has become even more pronounced with the widespread use of large language models, which have made fake reviews harder to detect. However, existing spammer group detection algorithms have certain limitations. For example, they often overlook the core–periphery structure within spammer groups, failing to adequately focus on the core reviewers who play a crucial role in group operations. Additionally, these algorithms struggle to detect spammers who are active across multiple groups. To address these challenges, we propose an overlapping spammer group detection algorithm based on key node propagation (KNP-OSG). First, we model the review data as a co-review graph and use the Deep Q-Network algorithm combined with an action filtering mechanism to identify key reviewers, or key spammers, who have a critical impact on spammer group detection. Subsequently, based on the structural relationships among pivotal spammers, an improved label propagation algorithm, copra-g, is proposed to further identify spammer groups. Experimental results show that the KNP-OSG algorithm outperforms existing methods on real-world datasets, demonstrating its effectiveness in detecting overlapping spammer groups.