Zero-Trust Access & Segmentation
← Identity & Access · ← Main Portfolio
Business problem
Flat networks — in hospitals, warehouses, or home labs — allow lateral movement from compromised devices to critical workloads. Access must be segmented by trust level with explicit policy at boundaries.
Constraints
- IoT and OT untrustworthiness — firmware-stale devices cannot share L2 domains with hypervisors or inference hosts
- Cross-subnet automation — orchestration must work across zones without flattening security
- Enterprise parallel — same principles apply when separating Baxter resilience layers from hospital OT, or warehouse telematics from ERP integrations
Architecture
Network trust zones
| Zone |
Trust |
Access policy |
| Perimeter / production |
Highest |
Hypervisors, NAS, AI inference, authoritative data |
| House / corporate LAN |
Medium |
Management, MQTT broker, wired endpoints |
| IoT / OT WLAN |
Lowest |
Smart devices — no route to core compute |
- MikroTik RouterOS — inter-VLAN firewall rules and NAT at gateway
- Traefik — deliberate ingress with TLS termination for exposed services
- SSID-to-VLAN mapping — WiFi clients land in correct trust zone by default
Credential-scoped service catalogs
- MCP tool servers — agents invoke only catalog-registered tools with environment-filtered credentials (Local AI & MCP)
- Backend catalogs — LiteLLM/Ollama backends filtered by available secrets and hardware capability
Tradeoffs
- Complexity vs security — multi-VLAN design adds routing and firewall maintenance; flat networks are simpler but unsafe
- Discovery vs isolation — mDNS repeaters enable media/IoT discovery without merging trust zones
Outcome
- Architecturally blocked lateral movement — compromised IoT cannot reach Proxmox or Ollama hosts
- Production-faithful sandbox — patterns pressure-tested locally before prescribing to healthcare and enterprise clients
- ALSTOM parallel — same OT/IT segregation mindset in industrial interoperability case study
Related deep dive → Infrastructure & Home Lab