Manufacturing Network Segmentation Guide
A plant network can fail long before a controller stops communicating. A ransomware event on a business workstation, an unmanaged remote-access connection, or a compromised engineering laptop can create a path into production systems that were never designed to withstand enterprise-level threats. This manufacturing network segmentation guide explains how to reduce that exposure without creating unnecessary operational complexity.
Thank you for reading this post, don't forget to subscribe!Network segmentation is not simply an IT firewall project. In a manufacturing environment, it is a controls architecture decision that affects PLC communications, SCADA visibility, historian data, vendor support, alarm response, safety functions, and plant uptime. The objective is to limit the consequences of a fault or cyber incident while preserving the deterministic, reliable operation required on the plant floor.
Why manufacturing network segmentation matters
Flat networks are common in aging facilities. An operator workstation, a packaging PLC, a quality database, a remote VPN connection, and an office computer may all share broad network access because the system expanded over years of incremental upgrades. The arrangement may work under normal conditions, but it provides few boundaries when something goes wrong.
Segmentation creates controlled zones and defined communication paths between them. Rather than allowing any device to communicate with any other device, the architecture permits only the traffic required for a production function. A historian may collect approved tags from a SCADA network. An engineering workstation may reach designated controllers during an authorized maintenance window. A business user may view production reports without gaining direct access to a PLC.
The operational benefits are significant. Properly designed segmentation reduces the attack surface, limits lateral movement, improves troubleshooting, and makes ownership of network connections more visible. It can also support compliance efforts under frameworks such as IEC 62443, particularly where asset identification, security zones, conduits, and controlled access are expected.
Segmentation does involve trade-offs. Restrictive rules can interrupt communications that production teams depend on if they are implemented without a complete understanding of the process. The answer is not to open every port after startup issues appear. The answer is to document required traffic, validate it in stages, and build an architecture that reflects actual plant operations.
Start with the process, not the firewall
A useful manufacturing network segmentation guide begins with an asset and communications assessment. Before defining zones, identify what is connected, what each asset does, and what can happen if it becomes unavailable or compromised.
Controls engineers and operations personnel should be involved from the start. IT may understand address management, identity controls, and enterprise risk, while plant personnel understand which HMI controls a critical line, which PLC has a legacy protocol dependency, and which remote vendor connection is needed during an outage. Neither perspective is sufficient on its own.
Document the major assets across the environment: enterprise systems, plant business applications, historians, SCADA servers, industrial PCs, engineering workstations, PLCs, safety PLCs, remote I/O, network switches, drives, analyzers, and vendor-managed equipment. Record the operating system where applicable, network address, physical location, owner, firmware or software version, criticality, and required communications.
The communications inventory matters as much as the equipment list. Identify source and destination devices, protocol, port, direction of traffic, frequency, and operational purpose. A rule described as “allow SCADA traffic” is too broad to be defensible or maintainable. A rule that permits a specific SCADA server to communicate with specific PLCs using a defined industrial protocol is measurable and easier to troubleshoot.
Define zones by function and consequence
Manufacturing networks are often organized using the Purdue Enterprise Reference Architecture as a starting point. The model separates enterprise systems from site operations, supervisory control, basic control, and field devices. It remains useful, but it should not be copied mechanically. A modern facility may have cloud data services, OEM skids, wireless instruments, edge gateways, and remote-support systems that do not fit neatly into a simple layer diagram.
The practical goal is to group assets with similar security requirements, functions, and consequences of failure. Typical zones include the enterprise network, an industrial DMZ, site operations systems, supervisory control systems, basic control networks, safety systems, and vendor or remote-access services.
An industrial DMZ is particularly important where information must pass between IT and OT. It can host services such as replicated historians, patch-management staging, secure file transfer, remote-access jump hosts, and data brokers. This creates a controlled buffer instead of permitting direct connections from the enterprise network into a control zone.
Safety systems require special attention. A safety PLC, safety instrumented system, or emergency shutdown network should not be treated as just another controller subnet. The architecture must preserve functional independence, defined safety communication requirements, and the integrity expectations associated with the safety lifecycle. In process environments, segmentation decisions should be reviewed alongside IEC 61511 obligations, hazard analysis findings, and the facility’s established safety instrumented function design.
Build conduits around approved communications
Once zones are defined, create conduits between them. A conduit is the controlled path that permits necessary communication between two zones. Firewalls, managed switches, access-control lists, jump servers, authentication controls, and monitoring tools all contribute to enforcing that path.
The default position should be deny by default, then allow only documented traffic. That does not mean production personnel should be forced through an unworkable approval process for every troubleshooting task. It means temporary access should be deliberate, time-bound, logged, and removed when the work is complete.
Remote access is a frequent weak point. Vendor support should terminate in a controlled remote-access zone rather than providing a direct route to controllers or HMIs. Use named accounts, multifactor authentication, session logging where feasible, and a jump host that limits reachable systems. Maintenance teams should be able to support equipment efficiently, but remote sessions must not become permanent unmanaged pathways into the plant.
Engineering workstations deserve the same discipline. They often contain programming software, controller project files, device drivers, and credentials that can change production equipment. Isolate these workstations from general office use, control internet access, manage removable media, and define how updates are tested before deployment. A segmented network cannot compensate for an engineering laptop that moves freely between business systems and critical control assets.
Validate segmentation without risking production
Network changes should be engineered and commissioned like any other change to a control system. Start with a pilot area, a noncritical line, or a defined production cell when possible. Test the normal operating sequence, alarm delivery, historian collection, recipe transfers, reporting, remote support, and recovery procedures.
Testing must include abnormal conditions. Confirm what happens if the historian becomes unavailable, the enterprise connection drops, a firewall reboots, or an operator needs local control during a network outage. Critical control functions should continue safely when higher-level systems are unavailable. If production requires an enterprise server to remain online for a basic machine to run, the architecture has created an avoidable availability risk.
Develop a rollback plan before implementation. Record existing switch configurations, firewall rules, controller routes, and SCADA communication settings. Schedule cutovers around production constraints, provide on-site controls support for critical changes, and establish clear decision points for returning to the previous configuration. This is particularly important in plants with limited maintenance windows or high cost-of-downtime processes.
Maintain the architecture after startup
Segmentation degrades when changes are made without governance. A temporary firewall rule remains open after a service call. A new OEM skid is connected to the nearest available switch. A replacement HMI receives broad access because its old configuration is not documented. Over time, the plant returns to a flat network in practice, even if the original design looked sound.
Maintain current network diagrams, zone definitions, device inventories, and rule sets. Review remote accounts and temporary access regularly. Include network requirements in machine procurement specifications so new equipment arrives with documented ports, protocols, patch expectations, and remote-support methods. A lifecycle automation audit can identify obsolete assets, unmanaged connections, unsupported operating systems, and control-system dependencies before they become a production event.
Monitoring also matters. Log firewall events, remote sessions, and unusual traffic between zones. The goal is not to flood a maintenance team with alerts. It is to establish enough visibility to identify unauthorized pathways, repeated communication failures, and behavior that differs from the approved architecture.
A well-segmented manufacturing network gives operations a more controlled way to modernize. It allows data to move where it creates value while keeping critical controls, safety functions, and production assets behind purposeful boundaries. The best next step is usually a joint IT, OT, and safety review of one production area, followed by an implementation plan that improves security without compromising the process that keeps the plant running.







