Machine Builder Remote Access: Secure Vendor Connectivity

How is a machine builder’s remote maintenance access organized? Moving from a modem in the cabinet and a remote desktop tool on the HMI to approved, logged, time-limited access; architecture, approval flow, contract terms.
Technical summary
Technology
IDMZ, jump server, PAM, MFA, session recording, IEC 62443-2-4
Topic
Securely managing remote access by machine builders and integrators in OT
Level
Intermediate–advanced
Prerequisites
  • Basic knowledge of IT/OT segmentation
  • Experience with firewall rule management
  • Access to the OT asset inventory
In this article8 sections

The line has stopped, and the builder’s engineer wants to connect remotely. In most plants, this connection was set up long ago without IT and the security team knowing: a 4G modem placed in the cabinet, a remote desktop tool installed on the HMI, or a gateway permanently connected to the builder’s cloud portal. In this guide we explain how to find the existing uncontrolled paths and how to move them to an approved, logged, time-bound vendor access model.

Uncontrolled vendor access paths

Uncontrolled access falls into a few patterns. What they have in common is the same: they completely bypass the organization’s firewall and authentication infrastructure.

  • A cellular modem that came with the machine: During commissioning, the builder adds an industrial 4G router to the cabinet. The device goes out to the internet with its own SIM card and is connected directly to the machine network. The perimeter firewall never sees this traffic.
  • A remote desktop tool on the HMI or engineering workstation: A tool installed in unattended access mode is a permanently open door whose password sits with the vendor. Because the station is also connected to the OT network, an attacker reaches the PLC directly.
  • A builder cloud gateway: Devices that set up a permanent outbound tunnel to the builder’s platform. They’re easy to install; but you can’t see what authentication the tunnel uses or who uses it.
  • A permanent site-to-site VPN: A permanently open tunnel from the integrator’s office to the OT network. If the vendor’s network is compromised, your OT network becomes an extension of it.
  • Personal VPN accounts: Accounts given to the vendor over the corporate VPN, reaching OT with broad rules and left open indefinitely.

From an enterprise cybersecurity perspective, the risk of these paths isn’t only an external attack. A vendor engineer can connect to the wrong machine, load a program while production is running or change a parameter without the operator knowing. Each of these is a safety or quality incident.

Target model: access through an intermediary, approved and logged

In the target architecture, the vendor never connects directly to the OT network. The connection terminates on an intermediary in the industrial DMZ (IDMZ) between IT and OT; from there it continues only to the permitted target, only with the required protocol. This approach matches the recommendation in NIST SP 800-82 Rev. 3 for indirect access through a DMZ and IEC 62443’s zone–conduit model.

  1. The vendor connects to the organization’s remote access layer (VPN, ZTNA or a secure remote access platform) with a personal account and MFA.
  2. The access layer directs the user only to the jump server or PAM gateway in the IDMZ; it doesn’t give routes to OT subnets.
  3. Access from the jump server to the target HMI, engineering workstation or PLC is provided by a narrow firewall rule opened for the duration of the session.
  4. Throughout the session, the screen, commands and file transfers are recorded; the responsible operator on site knows about the session.
  5. When the time runs out, the account and rule close automatically.

The jump server itself is a critical asset. It shouldn’t be a member of the corporate domain, or it should sit in a separate management forest; it shouldn’t have direct internet egress, and only the required engineering software should be on it. A common mistake is the jump server turning over time into a general-purpose machine where everyone does everything. At that point it isn’t an intermediary but a new attack surface. For the general separation between IT and OT, see the IT/OT segmentation guide; here we focus on the vendor traffic passing over that separation.

Comparing access methods

There’s no single method that fits every plant. The table below compares common options in terms of security features.

Method Visibility Access scope Main limitation
4G modem on the machine None The entire machine network Bypasses the organization’s firewall
Remote desktop tool on the HMI Low The HMI and the connected network Always on, shared password
Direct VPN to the OT network Intermediate Depends on rules, mostly broad The vendor’s device enters OT
IDMZ jump server High Defined targets Recording and approval must be set up separately
PAM / secure remote access platform High Per target and protocol Licensing and operational overhead

In practice, most plants start with an IDMZ jump server. As the number of vendors and the frequency of sessions grow, they move to a PAM or an OT-oriented secure remote access solution that offers the approval flow and recording built in. When choosing the access layer, the ZTNA and VPN decision criteria article is useful; if you’ll use VPN, the enterprise VPN security controls apply in this scenario too.

Request, approval and time-limited session flow

Technical infrastructure alone isn’t enough; every session should be tied to a work order. A workable flow is:

  1. Request: Maintenance planning or the production manager opens a work order: which machine, what purpose, which vendor person, estimated duration.
  2. Operations approval: The line manager approves that the session fits the production state. If there’s a program upload or parameter change, the conditions for stopping the line are set here.
  3. Security approval: For routine work, it’s given automatically with a predefined template; for unusual targets, the OT security owner approves.
  4. Activation: The vendor account and the related firewall rule are opened only for the approved time window.
  5. Session: The operator on site knows the session has started; if needed, they watch it and end it.
  6. Closure: The vendor writes the changes they made into the work order, the account closes, and the changed PLC program is backed up and compared with the previous version.

Define a separate, fast approval path for emergency breakdowns. If you don’t, at the first breakdown on the night shift the teams will turn the modem you shut off back on. We’ve seen this many times. The emergency path shouldn’t stay off the record either; only the approval step should be completable afterwards.

Session recording and monitoring

Session recording is both a deterrent and the main source for post-incident investigation. The recording scope should include at least:

  • User, source IP, start and end time, target connected to
  • Screen recording for graphical sessions; command recording for SSH or Telnet sessions
  • File transfers: file name, direction, size and, if possible, hash
  • Additional connections opened during the session

Keep recordings not on the jump server but in a separate, immutable store. Whoever compromises the jump server also deletes the recordings on it. Send session open/close events to the SIEM and tie every login outside the approved time window to an alarm rule. If there’s a passive monitoring sensor on the OT network, match PLC write and program download events seen during the vendor session with the session; that way you tell within minutes whether the change was authorized. This monitoring setup is described in detail in the OT attack detection article.

Contract and technical requirements

Vendor access is as much a contract matter as a technical one. IEC 62443-2-4 defines the security capabilities that system integrators and service providers must meet and can be used as a reference in procurement specifications. The clauses we recommend adding to the contract or technical specification:

  • Access will be made only through the method provided by the organization; the vendor won’t install its own modem or remote access tool
  • The list of names of the people who will access, and the notification period for changes to this list
  • Accounts will be personal, won’t be shared and will use MFA
  • The vendor’s obligation to report security incidents it experiences on its own side
  • Devices used for connecting will be up to date and protected against malware
  • A written list of the ports, protocols and destinations the machine needs
  • Every change made will be entered into the work order, and the current program will be handed over

Put these clauses in the specification at the quotation stage when buying a new machine. Negotiating after commissioning is much harder; the machine is on site and you have no leverage left. The general cyber risk assessment of vendors, questionnaires and contract processes can be handled within vendor and third-party cyber risk management.

Inventorying and closing existing access

Find the existing paths before setting up the new model. The inventory is fed from three sources: a physical site walk, network traffic and endpoint review.

Site walk

Look for antennas, SIM card slots or network devices you don’t recognize in machine cabinets. Ask the maintenance team and machine owners directly: “how does the builder connect when there’s a fault?” The answers often reveal paths not in the documentation.

Network traffic

On the perimeter firewall, review connections and DNS queries going from OT subnets to the internet. Regular traffic to the domains of remote desktop tools and builder portals is a sign of unattended access. If you’ve done passive asset discovery, unidentified routers on the OT network or devices using a second default gateway also become visible this way; for the method, see the passive asset discovery in OT networks article.

Endpoint review

On Windows-based HMIs and engineering workstations, list installed remote access software and persistent outbound connections. The PowerShell commands below collect information locally without changing anything:

# Search installed applications for remote access tools
$keys = 'HKLM:\SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall\*',
        'HKLM:\SOFTWARE\WOW6432Node\Microsoft\Windows\CurrentVersion\Uninstall\*'
Get-ItemProperty $keys -ErrorAction SilentlyContinue |
  Where-Object { $_.DisplayName -match 'TeamViewer|AnyDesk|VNC|RustDesk|Splashtop' } |
  Select-Object DisplayName, DisplayVersion, InstallDate

# Related running services
Get-Service | Where-Object { $_.DisplayName -match 'TeamViewer|AnyDesk|VNC' } |
  Select-Object Name, Status, StartType

# Established connections and the owning process
Get-NetTCPConnection -State Established |
  Select-Object LocalAddress, RemotePort, RemoteAddress,
    @{n='Process';e={(Get-Process -Id $_.OwningProcess -ErrorAction SilentlyContinue).ProcessName}}

Make a decision for every path you find: remove it, move it to the organization’s intermediary, or accept it temporarily and apply a compensating control. Plan the closure with the machine owner and don’t tear out the old path before testing the new one. An access method that doesn’t work at the next breakdown means the uncontrolled path comes back.

Running this transformation as part of the OT security program makes it consistent with inventory, monitoring and incident response; for a comprehensive approach, take a look at the OT/ICS and industrial cybersecurity service. Two steps are enough to get started: list the existing paths, and move the vendor that connects most often to the jump server model first. The main external reference for OT remote access architecture is NIST SP 800-82 Rev. 3.

Frequently asked questions

What should be done if the machine builder won’t accept removing its own 4G modem?

If the modem stays, at least make it disconnectable with a removable key switch or a contactor, keep it off normally and turn it on only for the duration of an approved session. Ask in writing which network the modem connects to, which ports it opens and who owns the cloud account. When the contract is renewed, make access through the organization’s intermediary a requirement.

Why is using a shared account for vendor access a problem?

With a shared account you can’t know who opened the session; access for someone who left the vendor continues unnoticed, and when the password changes the whole service is disrupted. When a personal account, MFA and session recording are used together, both accountability and incident investigation become possible.

How long should session recordings be kept?

There’s no fixed mandatory period; the period is determined by your incident investigation needs, the vendor contract, your internal audit cycle and whether the recordings contain personal data. Tie the retention period to a written policy, keep the recordings in an immutable store and delete expired ones regularly.

Does a ZTNA product meet this need on its own?

No. ZTNA improves authentication and application-level access; but in OT, the approval flow, notifying the operator, session recording and access to the target device only with the required protocol must be designed separately. Verify one by one whether the product you choose provides these functions.

Support

Contact us for support or a discovery session on this topic.

Contact form
[email protected]

Eretek Cybersecurity

Firewall, VPN, endpoint security and threat prevention. We reduce the attack surface and measure and report the risks.

Knowledge Center

Other articles