Graph fraud detection (GFD) on transaction networks is crucial for safeguarding financial systems. However, due to the limited perspective of existing graph neural networks (GNNs) in the single transaction view, sophisticated fraudsters can disguise themselves to exhibit weak fraud signals, appearing as borderline fraudsters. To address this challenge, we propose MH-LGC, a multi-view hypergraph fraud detection model with large language model (LLM) guided contrastive learning. MH-LGC tackles two key limitations of existing GNN-based GFD methods: (1) Due to the local aggregation mechanism, existing methods struggle to capture high-order trading patterns among distant fraudsters. MH-LGC introduces two temporal hyper-views as complements to the transaction view and employs a Temporal Hypergraph Attention Network (THAN) to integrate the three views. (2) Most GFD methods overlook the rich semantic cues embedded in transaction data. Although some general graph learning studies have explored LLM integration, the high computational overhead and task-specific fine-tuning make them impractical for GFD tasks. MH-LGC introduces a semantic view through a fine-tuning-free LLM-Guided Contrastive learning (LGC), adopting a novel paradigm for integrating GNN and LLM to reduce the computational overhead of LLM. Extensive experiments on three real-world datasets demonstrate that MH-LGC outperforms twelve state-of-the-art baselines, with AUC improvements ranging from 1.10% to 5.70%.

Fraud detection on streaming transaction graphs (STGs) faces challenges on the catastrophic forgetting of previously learned fraud patterns when adapting to evolving patterns. Although some Graph Continual Learning (GCL) approaches mitigate this issue by storing and revisiting historical samples, practical storage constraints prevent them from fully preserving previous patterns. In this work, we propose STG-DGR, a streaming GNN model with diffusion-based generative replay that generates synthetic samples to retain previously learned patterns without storing real samples. The generation of replay samples for STGs faces two key challenges: (1) Heterogeneity challenge of generating STG samples with discrete adjacency table, user features, transaction features, and transaction timestamps. (2) Dependency challenge of capturing bottom-up dependencies across layers in STG samples. To address these challenges, STG-DGR integrates two novel components: (1) a Computational Subgraph Processor (CSP) that transforms heterogeneous STG samples into well-organized hierarchical subgraphs, and (2) a Diffusion-based Subgraph Generator (DSG) that captures the bottom-up dependencies using a novel Transformer-based Hierarchical Denoising Network (THDN), and generates synthetic replay samples that preserve these dependencies. Extensive experiments on four streaming fraud detection datasets demonstrate STG-DGR's superiority in reducing forgetting and improving accuracy over nineteen state-of-the-art baselines.

Fraud detection on dynamic graph (FDDG) is in high demand across many real-world applications, such as financial transaction networks and social networks. A graph neural network (GNN) is an advanced methodology for learning graph data and has been widely adopted in fraud detection tasks. Despite the promising results, research on GNN-based fraud detection faces two critical challenges. First, due to the lack of temporal consistency mining in existing methods, camouflaged fraudsters can easily evade detection by imitating the behavioral patterns of benign users. Second, with constantly evolving fraud patterns and a scarcity of fraud samples, existing methods can easily overfit to current fraud patterns and struggle to identify the evolved and new ones. To address these challenges, we propose CamFD, a semi-supervised camouflage-aware fraud detection model on dynamic graphs. To detect the temporal inconsistency in the behavioral patterns of camouflaged fraudsters, CamFD is equipped with a novel consistency-sensitive contrastive learning (CSCL) module. CSCL discriminatively learns the consistency and inconsistency in users’ behavioral patterns by constructing temporal and structural contrastive pairs. Since modeling the evolving fraud patterns with sparse fraud samples is almost impractical, CamFD concentrates on modeling the relatively stable patterns of extensive benign users with a multivariate Gaussian distribution modeling (MGDM) module. We conduct extensive experiments on a private credit card fraud dataset as well as three public datasets. The experimental results demonstrate that CamFD outperforms ten state-of-the-art (SOTA) baselines across all datasets, with the area under ROC curve (AUC) improvements ranging from 1.03% to 5.32%.