Reconnaissance in the Age of AI: Exploring Modern ML Infrastructure

Modern AI deployments expose a broad and interconnected attack surface — including inference servers, vector databases, model registries, orchestration platforms, notebooks, and object storage — and this report details how adversaries can fingerprint, enumerate, and exploit misconfigurations across these components. It describes common fingerprints, ports, endpoints, gRPC and metrics enumeration, MLflow and vector DB enumeration, supply-chain risks (e.g., exposed HF tokens, dependency confusion), and operational mitigations such as VAPT, red teaming, and managed detection.