Cycle-Consistent Account Matching from Profile and Interaction Hypergraphs

Abstract

The proliferation of digital platforms has resulted in fragmented user identities distributed across multiple social networks. Reconstructing a unified user identity through account matching is crucial for applications ranging from personalized recommendation systems to cybersecurity. Traditional methods primarily rely on pairwise linkage using dyadic graphs, which fail to capture the complex, higher-order relationships inherent in modern social interactions. Furthermore, when extending pairwise matching to multiple platforms, existing models suffer from error propagation and lack global structural coherence. To address these limitations, this paper proposes a novel framework that leverages both profile and interaction hypergraphs while enforcing cycle consistency across multiple networks. By representing multi-way user interactions and shared profile attributes as hyperedges, the proposed approach captures intricate topological structures that dyadic graphs overlook. Additionally, the integration of a cycle-consistency optimization mechanism ensures that matches across three or more platforms remain logically coherent, significantly reducing false positive alignments. Through extensive descriptive analysis and algorithmic design, this study elucidates how hypergraph convolutional networks can be adapted for multi-network alignment without relying on mathematical formalizations. Comprehensive evaluations on simulated and real-world multi-platform datasets demonstrate the superiority of the proposed framework over state-of-the-art baselines in terms of precision, recall, and global structural integrity.