Back to all posts
Control Systems7 min read·November 1, 2026

How Should You Separate Machine and Office Networks?

TB
ThynkBlox Automation Team
Automation Engineering

To separate a factory machine network from an office network, define which equipment belongs together and put controlled boundaries between those groups. Inventory the communications production needs, then permit those specific connections through agreed firewall rules. Before changing the live network, test production dependencies, engineering access and rollback arrangements. Treat segmentation as one part of plant protection, alongside access control, recovery planning and maintenance.

Where should the network boundaries go?

Start with the production process, its dependencies and who needs access. Avoid making the office network and the entire factory floor the only groups you consider.

Draw the proposed boundaries around equipment with a shared operating purpose. A machine controller, its operator panel and its associated devices might belong together. A separate production line, shared supervisory system and engineering access service each deserve an explicit placement decision.

For every boundary, answer:

  • Which equipment and functions belong inside it?
  • Which connections must cross it?
  • Who owns the equipment and approves access?
  • What should production do when a permitted connection becomes unavailable?

Keep essential machine communications within the machine boundary where practical. Document any dependency that crosses into shared plant services or office systems.

Specify where traffic will be enforced and inspected. Do not accept different network names, address ranges or VLAN assignments as the acceptance evidence by themselves. Require tests showing that approved traffic passes and prohibited traffic is blocked.

How do you inventory the traffic production actually needs?

Build a communication register before drafting firewall rules. Combine available drawings, configurations, approved traffic observations and discussions with operators, maintenance staff and system suppliers.

Begin with the approach in surveying an industrial network before expansion. For this project, extend the survey into a list of connections with a clear operational purpose.

For each connection, record:

  • The source and destination equipment.
  • Which endpoint initiates the connection.
  • The protocol, destination port and any additional connection requirements.
  • The production function it supports.
  • When it runs, including occasional maintenance activities.
  • The expected behaviour if communication fails.
  • The owner who can confirm whether it is necessary.

Do not turn every observed connection into a permanent permission. Ask whether it supports an agreed function or belongs to an obsolete application, temporary tool or previous installation.

Include activities outside normal production: recipe transfers, backups, programme downloads, time synchronisation, authentication, name resolution and licence checks. Mark uncertain dependencies for investigation before cutover.

Agree an observation method suitable for the equipment. Do not introduce active discovery or diagnostic traffic onto the live machine network without assessing the method with the responsible controls specialist and equipment supplier.

What should the firewall rules allow?

Write each rule around an approved communication requirement. Start with unapproved traffic blocked, then add the permissions needed for documented functions.

The following is an illustrative rule-planning example, not a configuration to copy.

Requested connectionProposed decisionEvidence required
Office reporting application to plant reporting servicePermit the agreed reporting interfaceReports work; controller access remains blocked
Plant data collector to named machine controllersPermit required collection trafficRequired values arrive; unrelated connections fail
Approved engineering workstation to a controllerPermit through the controlled engineering routeAuthorised maintenance works; direct office access fails
Vendor laptop to the entire machine networkReject the broad request; define specific work and destinationsOnly the approved equipment and functions are reachable
Machine equipment to unrestricted internet destinationsLeave blocked unless a specific requirement is approvedEach exception has an owner, purpose and test

For each permitted flow, specify the initiating endpoint, destination, service, reason, owner and review condition. Where practical, define named equipment rather than broad network ranges.

Distinguish session initiation from return traffic. Ask the firewall designer to document how the proposed configuration handles replies and protocols with additional connection requirements.

Avoid making a temporary “allow everything” rule the commissioning baseline. If an exception is necessary, record its scope, expiry, approver and removal test.

How should engineering and vendor access work?

Design maintenance access before removing existing routes. Give engineers a defined path that supports the work they are authorised to perform.

Consider an approved engineering workstation or jump host, named user accounts, strong authentication, restricted destinations and access limited to the maintenance window. Decide how the route is enabled, monitored and closed.

Test the actual engineering tasks: connecting to the controller, viewing diagnostics, transferring an approved programme, obtaining a backup and restoring access after a workstation restart. A successful login alone is insufficient acceptance evidence.

Review laptops and workstations with multiple network connections. Require approval for each connection and verify that no unintended route bypasses the boundary.

For external support, follow the same ownership and approval process described in controlling vendor remote access to plant systems. Define who authorises the session, what the vendor may reach and who confirms that access has ended.

Include an emergency access procedure. Assign responsibility for enabling it, recording its use and restoring normal restrictions afterwards.

Which dependencies need testing before a live change?

Turn the communication register into a test plan. For every required connection, include a normal-operation test, a communication-loss test where appropriate and a recovery test.

Check dependencies across the full operating cycle, including:

  • Machine start, normal running, controlled stop and restart.
  • Operator commands, alarms and supervisory displays.
  • Order receipt, recipe selection and production reporting.
  • Data collection, timestamp alignment and buffered-data recovery.
  • Engineering diagnostics, backup and approved software transfer.
  • Shared services such as authentication, name resolution and time synchronisation.

For each test, define the expected result and the person who will witness it. Check the outcome at the receiving application or machine, not just whether a network connection opens.

Explicitly test prohibited paths too. Confirm that an office workstation cannot reach restricted controller services and that a maintenance account cannot reach unrelated equipment.

Use an isolated test environment or a controlled pilot where practical. If a representative test cannot be completed before cutover, record the limitation and agree how it will be tested safely during the change window.

Review any safety-related communication or operating dependency with a qualified assessor before changing it. Keep safety acceptance separate from ordinary connectivity checks.

How should the live cutover be controlled?

Prepare a written change sequence with named responsibilities. Save the relevant network and device configurations, and confirm how they will be restored.

Agree the stop conditions before starting. Examples include loss of an essential operator command, unavailable alarms, failed order transfer or loss of the approved engineering route.

The guide to defining rollback criteria for a controls cutover explains how to make these decisions concrete. Apply those criteria to the production functions affected by the network change.

During cutover:

  1. Confirm the plant is in the agreed operating state.
  2. Apply the approved boundary and rule changes.
  3. Run the production and blocked-access tests.
  4. Review denied traffic against the communication register.
  5. Restore the previous configuration if rollback criteria are met.

Do not approve broad permissions simply to clear firewall logs. Investigate each unexplained dependency and document the decision.

What common mistakes should you avoid?

Treating the network drawing as proof. Require evidence from both permitted and prohibited connection tests.

Observing only a normal shift. Include infrequent maintenance, restart, backup and recovery activities in the inventory.

Giving engineers unrestricted access. Define the equipment and tasks each access route supports.

Keeping temporary exceptions indefinitely. Give every exception an owner and a removal condition.

Declaring the plant secure after segmentation. Continue reviewing accounts, software changes, recovery arrangements and physical access. Describe the specific boundary controls that were tested, rather than making a general security claim.

What questions come up most often?

Must every machine have a separate network?

Choose boundaries according to operating dependencies, ownership and the consequences of lost communication. Document why equipment is grouped together. Avoid creating boundaries that the plant cannot maintain or test properly.

Can office users still see production reports?

Plan a defined reporting interface with an approved destination and service. Test that users receive the information they need while restricted machine services remain inaccessible.

What if an old machine needs a broad firewall exception?

Ask the supplier to identify the exact communication requirements. Test a narrower rule set where practical. If a broad exception remains necessary, document its exposure, compensating controls, owner and replacement plan.

When is the separation work complete?

Close the work when approved flows operate correctly, prohibited paths fail, recovery tests pass and temporary exceptions are resolved or explicitly owned. Hand over the final diagram, communication register, rule set, access procedure and test results to the people responsible for maintaining them.

Next step

Want to apply this to your plant?

Every project starts with a fixed-fee assessment, and you can stop after the report.