In the fast-paced world of financial services, traditional fraud detection methods can’t keep up with increasingly sophisticated fraud tactics and the sheer volume and complexity of transactions.
An AI-driven transaction monitoring solution stack has evolved that combines Graph Neural Networks (GNNs) and Federated Learning (FL), enabling organizations to rapidly deploy scalable, privacy-preserving, and secure fraud detection systems.

We’re excited to announce the release of a new OCI Landing Zone Workload Template that enables easy deployment of this cutting-edge solution stack for AI-driven transaction monitoring on OCI’s GPU infrastructure. The solution utilizes NVFlare for federated learning, with GraphSAGE inductive GNN model, implemented via PyTorch Geometric to deliver advanced anomaly detection and classification.


What are landing zones:

OCI Landing Zones are well-architected, configurable Terraform automation templates designed for various use cases. Landing Zones accelerate time-to-production on OCI and enable an optimal cloud environment that is secure, resilient, and cost effective.   Landing Zones’ templates, called Blueprints, are built using the common modules of the OCI landing zones framework. The framework and all templates are offered free of charge on GitHub.  Customers can select, configure, and deploy landing zones via GitHub, CLI, CI/CD, the OCI console, or using Resource Manager or Fleet Application Management services.   
The OCI Core Landing Zone template offers a general-purpose blueprint for 1-click provisioning of an optimized architecture and hardened configuration for tenancies and workload-ready infrastructure – deploying essential cloud services across security, identity, networking, and observability.   
Before deploying specific workload templates—such as the new AI Transaction Monitoring template—customers first provision an OCI tenancy that’s ready for workload expansions by deploying the OCI Core Landing Zone or Operating Entities Landing Zone. They then proceed to deploy workload templates.

What gets deployed by the AI Transaction Monitoring Workload template:

The OCI Landing Zone workload will configure the infrastructure services, deploy a Compute with GPU Shape, and create a GNN transaction classification model using GraphSAGE and Elliptic++ transaction dataset.  In this template, a single GPU-based compute instance is deployed, optionally with a dedicated application load balancer and backend set. 
The prerequisite resources listed below must exist prior to deploying this workload. You can make sure these resources are provisioned in your tenancy during the deployment of the OCI Core Landing Zone by selecting the appropriate settings when prompted during the deployment.
•    A workload compartment for holding the compute instance and block volume.
•    An application compartment with a private application subnet with a VCN, including a DRG and NAT gateway for outbound access to the internet.
•    Optionally, a public web subnet for a load balancer used in a mesh network environment.


Conclusion

As financial fraud becomes more sophisticated, so must our defenses. The integration of Graph Neural Networks with Federated Learning, powered by GPUs on OCI, offers a cutting-edge, privacy-aware, and collaborative method for real-time, AI-enabled transaction monitoring.
Through automation, visualization, and a scalable deployment model, OCI Landing Zones equip organizations to provision an optimal cloud tenancy and deploy this AI workload quickly, helping ensure a secure, resilient architecture for streamlined operations. 


To get started today, simply download and deploy one of these landing zones and the workload template: 
OCI Core Landing Zone or Operating Entities Landing Zone
AI Transaction Workload template