IT/OT Segmentation in Manufacturing Plants

In what order are IT and OT networks separated? Mapping existing flows, an industrial DMZ, and firewall rules monitored first and enforced later, with a production impact assessment at every step.
Technical summary
Technology
IEC 62443, industrial DMZ, Purdue model
Topic
Industrial security
Level
Intermediate
In this article3 sections

If ransomware that starts with an email attachment opened in the office can reach the operator panels on the production line within a few hours, the problem isn’t the malware but the network architecture. In many manufacturing plants, office computers, servers, PLCs and operator panels sit on the same network or on freely routed VLANs. IT/OT segmentation is the basic step to reduce this risk. The sequence below is built on assessing the production impact of each change before implementing it, because what can make segmentation projects stall halfway is the line stopping on the first attempt. To implement the zones and the migration sequence on site, see our OT network segmentation implementation in the factory service.

Reference model: Purdue levels and IEC 62443 zones/conduits

The Purdue model is a common reference that defines the layers from the corporate network (levels 4–5) down to sensors and PLCs in the field (levels 0–1). IEC 62443-3-2 recommends dividing the system into zones with similar security requirements and the communication paths between these zones into conduits, and setting a target security level for each. The two approaches can be used together.

In practice, Purdue gives you the “where am I” map, and the zone/conduit model the decision of “what will I protect, against what, and how much.” Don’t try to turn Purdue levels one-to-one into VLANs. In modern plants, cloud-connected sensors, machines writing data directly to the ERP and wireless handheld terminals already blur the layers. Defining a zone as assets with the same risk profile and the same owner gives healthier results.

Step-by-step segmentation

1. Passive discovery and flow map

OT devices can be sensitive to active scans. That’s why the inventory is built by passively analyzing traffic taken from a SPAN/mirror port on the switch. The result documents which IT system talks to which OT system over which protocol (e.g. ERP ↔ historian, engineering workstation ↔ PLC, remote maintenance ↔ HMI).

A network scan that’s routine on the office network can hang an old PLC’s communication module. This risk isn’t theoretical. During discovery, use active tools only with the approval of the production and maintenance team, during a planned shutdown and against individual devices.

Writing only IPs and ports in the flow map isn’t enough. For each flow, the source, destination, protocol (such as Modbus/TCP, EtherNet/IP, S7, OPC UA), direction, frequency and business owner should be recorded. To avoid missing infrequent flows, keep the observation long enough to cover shift changes, weekend maintenance and recipe uploads.

2. Target architecture: an industrial DMZ

An industrial DMZ (level 3.5) is placed between IT and OT. The principle is simple: no session passes directly from IT to OT. Data transfer happens through intermediate systems in the DMZ.

DMZ component Purpose
Historian replica / intermediate server Transfer of production data to ERP and reporting systems
Patch and antivirus distribution server Controlled updates to OT systems
Remote access gateway Vendor maintenance done with MFA, time-limited and logged
File transfer point Recipe and program files scanned before being transferred

For the DMZ to work, its two sides must terminate on separate firewall interfaces (or two separate firewalls), a connection coming from the IT side must stop in the DMZ, and a new connection to the OT side must be opened only from the system in the DMZ. It’s common to see a server in the DMZ connected to both the IT and OT networks with two network cards. That isn’t a DMZ; it’s a passage that bypasses the firewall.

Data direction is also a design decision. In most cases, production data needed for reporting can be pushed from OT toward the DMZ; systems in IT pull the data from the copy in the DMZ. That way not a single session initiated from IT toward OT is needed. Integrations that set up the reverse direction (such as the ERP reading data directly from a PLC) should be the first target of segmentation.

3. Firewall policy: monitor first, then block

At first, the firewall at the IT/OT boundary only logs. A few weeks of observation verify the real flows, production and IT approve them together; then traffic outside the allowlist is blocked during a planned maintenance shutdown.

At the moment of blocking, a rollback plan must be ready. Take a configuration backup before the rule change, keep the line supervisor on site and define in advance the “it works” criteria for each critical line (is the HMI receiving data, is the historian updating, can recipes be loaded). If something goes wrong, the last blocking rule added is rolled back, the flow is examined and it’s tried again at the next shutdown.

On firewalls that understand industrial protocols, rules can be written not only at the port level but also at the command level; e.g. giving the reporting system read-only access while blocking write commands to the PLC. This feature varies by vendor and license.

4. Separate shared services

IT and OT sharing the same Active Directory, file shares or management tools opens a path that indirectly bypasses the firewall. A separate identity domain or local administration and separate management workstations should be considered for OT.

When a shared domain is compromised, the attacker holds valid credentials on OT servers without being caught by segmentation rules. The same logic applies to monitoring and remote management tools: if the central management console in IT can send commands to OT devices, that console is part of OT and should be protected as such.

5. Cell and area separation

Additional separation between critical lines or machines from different vendors limits the spread of an incident on one line to the others.

Don’t lose manageability in cell-level separation. Putting a separate firewall on every machine isn’t sustainable in most plants. One zone per line and, where a machine vendor brings its own network, a small cell gateway working with NAT and a limited allowlist is usually a balanced start.

Common mistakes

  • Treating VLAN separation alone as enough; if routing between VLANs isn’t inspected, there’s no separation.
  • Forgetting, in the inventory, connections set up by machine vendors with their own modems or remote desktop tools.
  • Enabling rules in blocking mode without observation.
  • Not coordinating changes with production planning.

Two more common items can be added to this list. The first is “any-any” rules opened temporarily in the rule base becoming permanent. Write an end date and an owner on every broad rule opened during commissioning. The second is dropping monitoring after segmentation is complete. Flows change as new machines, new vendors and new integrations arrive. Firewall logs and the flow map should be compared at least every three months.

There’s also the question of ownership: who approves a change on the OT network? If this question isn’t answered before the project, every rule request is left hanging between IT and production. Define a simple change approval flow; the technical work speeds up after that.

To plan the work in your plant, see our IT/OT Network Segmentation page; for discovery and risk analysis, OT/ICS and Industrial Cybersecurity; and for a standards-based approach, IEC 62443 Consulting.

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