Remote OT Access on Vessels: Governing the Maker’s Backdoor Into Your Engine Room

Every engine maker, every propulsion vendor, every integrated bridge supplier now expects a way to reach their equipment on your ship remotely. If you have not designed that path, they have built it for you — and it is exactly the door an attacker wants.

The commercial logic is hard to argue with. A main engine control system throws a fault mid-ocean. The maker’s support engineer, sitting in a service centre thousands of miles away, can diagnose it in an hour if they can see the controller — or the ship waits days for someone to fly out. So the vendor asks for remote access, and somewhere on the vessel a cellular router, a spare satellite channel or a laptop with a modem quietly becomes a permanent tunnel into operational technology that steers, powers and stops the ship. Nobody governs it because nobody designed it. It just accreted.

This is a technology decision dressed as a support arrangement. The question is not whether vendors get remote access — they will, one way or another — but whether that access runs through a path you control, log and can switch off. What follows is a reference pattern for doing exactly that.

Free · 4 minutes

Would you survive contact with a determined attacker — or an auditor?

Fourteen questions on access, patching, detection, and recovery — the basics that prevent most real incidents, and the ones most often assumed rather than verified. Banded finding on screen, full sheet by email.

Why the ungoverned door is the worst one

The regulatory frame has caught up with the risk. IMO Resolution MSC.428(98) folded cyber risk into the ISM Code, so a ship’s safety management system has had to address it since the first annual verification of the company’s Document of Compliance after 1 January 2021. More concretely, IACS Unified Requirements E26 (cyber resilience of ships) and E27 (cyber resilience of on-board systems and equipment), adopted in 2023, apply to vessels contracted for construction on or after 1 July 2024. Both of them care a great deal about how remote access to onboard systems is controlled and recorded.

Strip away the instrument names and the substance is simple. An ungoverned vendor tunnel fails on every axis at once. You cannot say who used it, because there is no identity behind the shared credential the vendor rotates among its engineers. You cannot say what they did, because nothing recorded the session. You cannot bound where they went, because a flat network lets a support session into the fuel-injection controller reach the ballast system just as easily. And you cannot prove any of this to a class surveyor or a flag administration, because the evidence never existed. A standing connection into safety-critical OT with no identity, no recording and no segmentation is not a support channel. It is a pre-installed foothold.

The convergence of ship IT and OT has already dissolved much of the old air gap that people still assume protects the engine room. I have written before about what OT/IT convergence on modern vessels means for governance; remote maker access is the sharpest edge of that same problem.

The pattern: a broker, not a tunnel

The design principle is that a vendor never connects to your OT. They connect to a broker — a hardened jump host in a demilitarised zone between ship IT and ship OT — and the broker connects onward to a named piece of equipment, for a named window, under a recorded session. Nothing about the equipment is reachable from the outside directly. In IEC 62443 terms, the OT sits in its own zone, the DMZ is a separate zone, and the only conduit between them is the broker, policed by a firewall that permits precisely one flow.

Five properties make this a governed path rather than a nicer-looking tunnel.

  • Individual identity, not a shared vendor login. The support engineer authenticates as themselves through the broker. If the maker cannot name the human on the keyboard, the session does not open. This is the same access-governance bar that financial supervisors now hold firms to under regimes like NYDFS privileged access management — the maritime version is not different in kind.
  • Just-in-time, not standing, connectivity. The path to the equipment is closed by default. It opens on an approved request, for a bounded window, and closes automatically. There is no permanent route from the internet to the engine.
  • A destination allowlist. An open session reaches one target on one protocol on one port. The propulsion vendor’s engineer can reach the propulsion controller and nothing else.
  • Session recording that lands locally first. Every session is recorded — keystrokes, commands, screen — and written to storage on the vessel before any attempt to ship it ashore. A recording that only exists once the satellite link is good is a recording you will lose exactly when you need it.
  • Logging that tolerates a bad link. Access decisions and session metadata are logged at the broker and buffered locally, then synchronised to shore when connectivity allows. The audit trail cannot depend on the very link that is least reliable at sea.

What the broker configuration looks like

Below is a vendor-neutral policy for such a broker. It is illustrative, not a product configuration, but it shows the shape: a named vendor, an approval gate, a single allowlisted destination, local-first recording and store-and-forward logging.

# vendor-access-broker.yaml — governed OT support session policy
# Broker sits in the ship IT/OT DMZ. Nothing here grants standing access.

broker:
  bind: 10.20.0.10        # DMZ interface only; not reachable from OT zone
  recording:
    mode: full            # keystrokes + rendered session
    store_local_first: true
    local_path: /var/lib/otbroker/sessions
    forward_when_online: true   # ship ashore only after local write
  logging:
    sink_local: /var/log/otbroker/access.jsonl
    forward: store-and-forward   # buffer through satellite outages

access_rules:
  - vendor: "propulsion-maker-a"
    principals:               # individuals, never a shared login
      - engineer: "a.novak@maker-a.example"
    approval:
      required: true          # just-in-time; closed by default
      approver_role: "master_or_dpa"
      max_session_minutes: 120
    destination:              # single allowlisted target
      host: 10.30.4.21        # propulsion control gateway, OT zone
      port: 443
      protocol: https
    deny_all_other: true      # no lateral reach beyond this host

Read it as a specification the vessel and the maker both sign up to, not as a script. The firewall between the DMZ and the OT zone enforces the destination line; the broker enforces identity, approval and recording. If any layer is missing, the pattern degrades back into a tunnel.

Where fleets get this wrong

The most common failure is treating remote access as a per-ship afterthought rather than a fleet standard. Each vessel ends up with a different arrangement negotiated by whoever was aboard during the last drydock, and the shore security team has no consistent picture of who can reach what. The second failure is letting the maker supply and own the access appliance. If the vendor controls the jump host, the vendor controls the audit trail, and you are trusting the party you are meant to be governing to mark its own homework.

The third is forgetting that the pattern only holds if the OT zone is genuinely segmented from ship IT and from the crew and passenger networks in the first place. A broker in front of a flat network is theatre. The same zone-and-conduit discipline that shore-side regulators expect for industrial estates — the kind set out in OT and cloud cybersecurity control architectures — has to exist on the hull before a governed access path means anything.

The maker’s backdoor will exist whether you design it or not. The only real choice is whether it is a door you hold the key to, watch through and can bolt in an afternoon — or one someone else fitted, that you cannot see, and cannot prove you closed.

Most technology problems are not technology problems. They are control problems.

The systems exist. The investment has been made. The question is whether leadership can understand, direct, evidence, and sustain what those systems produce. Find out where control exists — and where it only appears to.