Resources

Federated Graph Intelligence for Financial Fraud Detection

A unified, privacy-preserving graph learning platform to combat financial fraud and illicit finance. This paper specifies a deployable technical design where financial institutions build person-to-person transaction graphs and train temporal graph neural encoders locally, then exchange signed, encrypted model updates via a consortium aggregator with robust aggregation and layered defenses against model poisoning and privacy attacks.

The official Python client for the Federated Graph Intelligence platform. Pull production model weights, train on local transaction graphs, and push updates back to the federation — all from a simple, Pythonic API.