AnthosGuard

• 

  AccountCreated (POST)

• 

  FraudDetector (GET)

Inspiration

Fraud is one of the fastest-growing threats in finance and digital platforms. Fake accounts and stolen identities can slip past static rules. We wanted to create a cloud-native, real-time fraud detection system that is scalable, intelligent, and easy for developers to integrate.

What it does

AnthosGuard detects fraudulent account creations in real time. • When a new account is created, an event is published. • Our fraud-detector service scores the account (LOW / MED / HIGH risk). • A live status API and dashboard display the latest detections. • Judges (or developers) can easily test it using a Postman collection with pre-configured environment variables.

How we built it

• Developed services locally with kind + Docker.
• Built microservices: account-created (event publisher) and fraud-detector (AI-based scoring).
• Deployed to Google Kubernetes Engine (GKE Autopilot) with Ingress and BackendConfigs.
• Used Artifact Registry for secure image storage and Workload Identity for service accounts.
• Created Postman demo scripts for an easy end-to-end test flow.

Challenges we ran into

• Docker installation conflicts on macOS.
• Migrating from local NATS to GCP Pub/Sub.
• Debugging Ingress 404/502 errors and load balancer health checks.
• Ensuring fast cluster recreation after accidental deletions.

Accomplishments that we’re proud of

• Running fraud detection live on GKE Autopilot with public endpoints.
• End-to-end event pipeline working reliably with real-time risk scoring.
• A judge-friendly demo flow via Postman with one-click reproducibility.
• Resilience: overcame infra and config issues under time pressure.

What we learned

• How to design and deploy cloud-native event-driven architectures.
• Best practices with GKE Autopilot, Workload Identity, and BackendConfig.
• The importance of making demos simple and reproducible for others.

What’s next for AnthosGuard

• Extend beyond account fraud to transaction-level anomaly detection.
• Add machine learning models for adaptive fraud scoring.
• Implement alerting & incident response integrations.
• Package AnthosGuard as a plug-and-play fraud detection module for fintechs, e-commerce, and SaaS platforms.

Built With

• artifact
• docker
• fastapi
• gke
• google
• kind
• kubernetes
• nats
• pub/sub
• pydantic
• python
• registry