Dual-scale graph transformers cut fraud and customer friction in banking
Dual-Scale Relational Graph Transformers for Ecosystem-Aware Fraud Detection
Machine LearningArtificial Intelligence
Summary
Fraud in digital banking is hard to detect because it depends on both small-scale details and the big-picture fraud patterns over time. The authors developed a new method called HERMES that looks at local connections around a transaction and also captures broader changes in fraud behavior across the whole system. Tested on millions of bank transactions, HERMES found fraud more accurately and caused less inconvenience to honest customers than existing systems. The method helps banks catch fraud while keeping genuine users happy.
What this means in practice
- •For digital banking fraud teams: Detect account takeover fraud more accurately by combining local transaction relations with global fraud ecosystem changes.
- •For financial data engineers: Improve fraud detection systems using dual-scale graph transformers to reduce false alarms and customer disruption.
Authors
Mohsen Nayebi Kerdabadi, Xinrou Li, Yao Xiao, Zijun Yao, Xin Sun
Abstract
Account takeover (ATO) fraud is a growing threat to digital banking, requiring effective detection while minimizing friction for legitimate customers. Production systems predominantly rely on tabular models that score sessions in isolation, discarding the relational structure of the underlying interaction network. Although graph-based models exploit relationships among sessions and network entities, they primarily reason over local neighborhoods and therefore capture only part of the problem: fraud risk depends jointly on the local relational structure surrounding a session and the evolving global state of the fraud ecosystem. We present HERMES (HEterogeneous Relational Micro--macro graph transformer Encoder for high-risk Sessions), a dual-scale architecture that jointly models these complementary scales of information. Micro-GT captures local heterogeneous graph structure through structured, relation-aware attention over a temporally safe session neighborhood. Complementing this local representation, Macro-GT models ecosystem-level context using non-anticipative climate tokens that summarize fraud dynamics, platform shifts, and infrastructure reuse, together with adaptive class prototypes that track representative fraud and benign session patterns over time. Evaluated on more than 130 million high-risk transaction sessions from a leading U.S. financial institution, HERMES consistently outperforms production and strong graph-based baselines, achieving a 44.44% relative reduction in customer friction and a 24.66% relative improvement in fraud recall over the production system. Ablation and temporal-stability analyses further demonstrate complementary gains from local relational modeling and global ecosystem context across changing fraud regimes.