Overview of the Financial Big Data Anti‑Fraud Technology Whitepaper

On December 1, 2022, China’s Anti‑Telecom Network Fraud Law came into effect, prompting Ant Group and Tsinghua University to release a whitepaper that details the infrastructure and core technologies for financial anti‑fraud systems.

The whitepaper identifies five key fraud characteristics—precursory signs, group behavior, correlation, concealment, and dynamism—and lists nine common scam types such as fake job offers, impersonation, and fraudulent shopping.

It proposes a three‑stage anti‑fraud approach (pre‑risk perception, in‑transaction detection, and post‑incident response) and argues that a complete financial risk‑control system must support all three phases.

At the infrastructure level, the paper recommends building a multi‑dimensional heterogeneous massive‑scale interaction graph to store, retrieve, and represent large‑scale, strongly linked, heterogeneous, and privacy‑sensitive data. Trustworthy AI is applied throughout model training, deployment, and operation to ensure data privacy, model robustness, interpretability, and fairness.

Pre‑risk control involves external threat perception, full‑network threat detection, and intelligence collection, with technologies such as risk content localization, qualitative risk perception, intelligence classification, and element extraction.

During‑risk control focuses on real‑time transaction risk identification and decision making, covering cold‑start challenges, structured/graph anomaly detection, full‑graph risk control, edge risk control, and decision strategies that balance risk and user experience using biased inference and interactive risk control.

Post‑risk control addresses user complaints (intelligent adjudication), missed‑fraud detection (case‑based retrieval), and risk‑control review (threat knowledge extraction), providing a feedback loop to refine pre‑ and in‑transaction defenses.

The whitepaper concludes with industry case studies demonstrating the practical application of these technologies and calls for continued collaboration between academia and industry to advance anti‑fraud ecosystems.