Security-aware and scalable community detection in multilayer social networks via semi-supervised matrix factorization

Detecting communities in multilayer social networks remains a complex challenge, particularly due to limited consideration of user attributes, inter-layer heterogeneity, and emerging security-related dynamics. To address these limitations, we propose an enhanced model called S3MFA (Secure and Scalable Semi-Supervised Matrix Factorization Algorithm), which integrates structural topology, user attributes, and inter-layer user overlaps into a unified nonnegative matrix factorization framework. Each network layer is independently processed using semi-supervised clustering guided by pairwise constraints, followed by a kernel-based fuzzy ensemble clustering strategy to generate a consistent global community structure. To further improve applicability in real-world environments, S3MFA incorporates a novel security-aware and scalable community detection module, which models trust dynamics, node vulnerabilities, and risk propagation. This enables the algorithm to uncover communities that are not only structurally meaningful but also resilient to security threats. Extensive evaluations across multilayer datasets confirm the superior performance of S3MFA over existing state-of-the-art methods in terms of accuracy, robustness, and scalability, especially under uncertain or adversarial conditions.