Kerberus
Real-time Solana transaction classifier and threat detector
§ overview
Streams on-chain activity through Yellowstone Geyser and runs anomaly detection at very low latency to flag suspicious transactions as they happen. Kerberus classifies transactions in real time — not after the fact — giving security and risk teams an early-warning signal on-chain.
kerberus.com ↗§ what we built
- →High-throughput Yellowstone Geyser streaming pipeline for real-time on-chain data ingestion
- →gRPC message processing for low-latency transaction classification
- →Anomaly detection engine in Rust identifying suspicious transaction patterns
- →Classification layer for known attack patterns, MEV signatures, and unusual account interactions
- →Alert system for downstream notification on flagged transactions
- →Production deployment with continuous stream handling and fault recovery
§ results
Kerberus classifies Solana transactions in real time, providing threat detection before confirmations settle rather than in post-hoc analysis. The system runs in production, processing the continuous Solana transaction stream and flagging anomalies as they occur.
§ tech stack
Need real-time on-chain monitoring?
We build streaming anomaly detection and classification systems for Solana and EVM chains — production-hardened and latency-optimized.