LOCUS
Neuro-Symbolic BGP Telemetry & Automated Defense platform.
Architecture & Overview
The internet's core routing protocol (BGP) is notoriously vulnerable to route leaks and malicious hijacks. I built LOCUS as a closed-loop architecture that doesn't just monitor global traffic, but actively defends against anomalies in real-time. The system utilizes a high-speed Go daemon tapping into the RIPE RIS Live firehose, feeding a dual-layer 'Neuro-Symbolic' engine. The Neuro layer uses an Unsupervised Machine Learning model (Isolation Forest) to flag statistical anomalies, while the Symbolic layer queries a Neo4j graph database to verify if the physical Autonomous System (AS) path is topologically possible. When a threat is verified, the FastAPI Python backend autonomously executes commands against a virtualized FRRouting network gateway to inject BGP blackholes. The entire Dockerized pipeline streams live telemetry via WebSockets into a dark-mode Next.js dashboard.