FalseRoute

Inspiration

Security teams need to respond quickly to suspicious activity, but experimenting with real production systems can be dangerous. We built FalseRoute to demonstrate a safer approach: use AI to analyze threats and recommend responses, while deterministic application rules keep every action bounded and explainable.

What We Built

FalseRoute is an agentic cyber-deception system that receives synthetic intrusion events, evaluates their risk, records the reasoning, and safely assigns suspicious activity to simulated decoys instead of touching production systems.

The project uses TypeScript, React, Express, PostgreSQL, Gemini, Pub/Sub, and Cloud Run. The dashboard shows the complete workflow—from event ingestion and AI analysis to policy decisions, simulated containment, and audit history.

What We Learned

We learned that autonomous security systems need more than intelligent recommendations. They also require strict validation, explicit authorization, durable audit records, safe failure handling, and clear provenance for every decision.

Challenges

The biggest challenge was balancing autonomy with safety. We designed a closed action catalog, deterministic policy enforcement, duplicate-event protection, bounded retries, and simulated effects so the system could demonstrate autonomous response without creating real-world risk.

Built With

• cloud-run
• docker
• express.js
• gemini
• google-cloud
• node.js
• postgresql
• prisma
• pubsub
• react
• terraform
• typescript
• zod