What is OT network segmentation?

OT network segmentation is the practice of dividing an operational technology network into isolated zones so that a compromise in one zone cannot spread to the rest of the plant. Each zone holds devices that share a function or a risk level, and traffic between zones is permitted by explicit policy rather than allowed by default. Done properly it contains an intrusion, limits how far malware can travel, and gives operators a defensible boundary between the corporate network, the industrial network, and the safety-critical systems at the edge.

That is the short version. The reason most OT networks are still flat is that OT is not IT.

Why OT segmentation is a different problem

In an IT network you can re-address subnets, push agents, patch on a schedule, and take a maintenance window. In OT most of those moves are off the table.

The devices cannot defend themselves, and many of them will never be patched. JP Morgan estimates that within the operational technology layer, 45% to 60% of hardware is patchable, 15% to 25% is unpatchable, and 25% to 35% has no patch available. The same research notes that operational infrastructure managing physical processes “often remains in service for 10 to 18 years,” against cloud systems replaced every four to five. You cannot install an agent on a fifteen-year-old PLC, so whatever segments it has to work without touching it. See protecting legacy OT without downtime.

Downtime is expensive and sometimes unsafe. Re-IPing a running process network or inserting new inter-VLAN firewalls usually means a planned outage. In continuous operations that outage has a dollar figure attached, and plenty of segmentation projects die because nobody can get the window approved.

The traffic is deterministic and brittle. Industrial protocols expect specific paths and specific timing. Segmentation that breaks a control loop is worse than no segmentation, so the approach that survives contact with a live plant adds isolation without changing the addressing or the physical topology underneath.

This is why flat networks persist, and why they are dangerous. A flat network means that anything which gets a foothold anywhere gets the whole plant. See what NotPetya did to Maersk.

The frameworks: Purdue Model and IEC 62443

Two frameworks shape how OT teams think about where the boundaries go.

The Purdue Model organizes an industrial environment into levels, from Level 0 field devices and sensors up through Level 3 site operations and into the enterprise levels, with a demilitarized zone between IT and OT. Segmentation maps onto those levels directly: isolate the field and control levels from site operations, and never let the enterprise reach down into control except through a controlled boundary.

IEC 62443 turns the idea into a formal method built on zones and conduits. A zone is a group of assets with a shared security requirement. A conduit is the controlled communication path between zones, and every conduit carries an explicit policy. The standard assigns security levels, target, capability, and achieved, so a team can state how much protection a given zone actually needs rather than over-segmenting or under-segmenting by instinct. For a practical walkthrough, see how much to segment your network to follow IEC 62443 and the IEC 62443 compliance mapping.

The frameworks tell you where the boundaries belong. They do not tell you how to enforce those boundaries on a live network full of devices you cannot touch. That gap is what the rest of this page is about.

OT network segmentation best practices

These hold across most industrial environments, from oil and gas to water utilities to manufacturing.

Inventory first, then group by function and consequence. You cannot segment what you cannot see. Cluster devices into zones by process function and by what happens if they are compromised, not by where they sit in the building.

Default-deny between zones. Traffic between zones should be blocked unless policy explicitly permits it. An allow-list of known, required flows is the entire point.

Plan for the assets your inventory missed. Every OT inventory is incomplete, and the honest question is what happens to the devices you do not know about. A default-deny posture answers it: an asset missed by manual entry, by import, or by a discovery scan is blocked the moment it tries to communicate and, where connection logging is enabled, the attempt is recorded.

Harden the IT/OT boundary. The route from the enterprise network into the plant is the most abused path into OT. Treat it as a controlled conduit, not an open door.

Scope remote and third-party access to one machine. An OEM technician needs a single device for a defined window, not a route into the network. See OT secure remote access.

Microsegment the high-consequence assets. For safety-critical systems, go past zone boundaries and isolate individual devices or small groups, so a foothold has nowhere to travel.

Know what is moving east to west, and know that segmentation is not the thing that tells you. Visibility and enforcement are different jobs. BlastWave works with Nozomi Networks, Claroty, and Dragos. Their platforms find what is on your network. An export from any of them can be imported into BlastShield as the baseline for a deployment, so the assets they discovered become the assets that stop answering.

Segment without stopping the process. Choose an approach that does not require re-IPing, new physical firewalls, or an outage on a running line. That is the difference between a segmentation project that ships and one that lives on a slide.

Traditional segmentation, microsegmentation, and overlay approaches

Not all segmentation is equal, and the differences matter most in brownfield plants.

VLANs and inter-zone firewalls are the traditional route. They work, but they mean re-architecting the network: new addressing, new hardware, ACL maintenance, and usually an outage to cut over. Greenfield, that is fine. In a plant built over twenty years it is slow, expensive, and risky. It is also slow to change afterward. A firewall or VLAN change moving through OT change control takes weeks.

Agent-based microsegmentation shrinks the boundary from the zone to the individual workload, so even devices inside one zone cannot reach each other freely. It limits lateral movement well, and it assumes you can install software on the endpoint. A PLC cannot run an agent, which means agent-based microsegmentation cannot enforce policy between two devices that cannot host one.

A software-defined overlay enforces policy in front of devices that cannot host anything. BlastShield supports an SDN-like Layer 2 isolation mode in which the switch forwards traffic to BlastShield, which enforces policy across Layer 2 switched segments inside the same broadcast domain. That is what makes east-west enforcement possible between two un-agentable devices on one switch. Added latency is near switch levels, microseconds rather than milliseconds, which is what makes it viable in front of cycle-time-sensitive traffic. Jitter is hardware-dependent and varies by hardware revision, so rather than publish a number that will not apply to your equipment, we measure it on yours during validation testing.

Policy activation runs in seconds, often under one second, from activation to enforcement. Against weeks through change control, that is the difference between containment as a plan and containment as an action.

What “hidden” actually means here

Multiple vendors now market asset hiding, so the word on its own tells a buyer nothing. The question worth asking any vendor is what the concealment is a property of.

BlastShield drops unauthenticated packets silently. Not refused, not reset, not returned with an error. Nothing is sent back, so a scanner cannot distinguish a protected address from empty address space. Reconnaissance, the first step in every documented OT attack chain, returns nothing to act on.

And the policy binds to the device, not to a location in the network. Under Layer 2 segregation the device is identified by its MAC address, an identifier intrinsic to the device, so if the device moves the policy moves with it. With the BlastShield Agent the policy is implemented directly on the device. Either way, protection travels with the asset rather than with a box it happens to sit behind, so moving an asset does not mean re-provisioning a position.

Both approaches reduce discoverability. Only one makes the identity of the protected thing the device itself.

A fair question: MAC addresses can be spoofed. They can. But look at what that requires. To spoof a MAC an attacker first has to discover the current device’s MAC and replace it, and that requires Layer 2 access to the segment. An attacker with Layer 2 access is already inside, which is an insider problem rather than an external one, and insider attacks are genuinely hard to defend against, ours included. The argument on this page is about external reconnaissance, the kind AI has made free, continuous, and universal. Moving the adversary from anyone with a scanner to someone already on your Layer 2 segment is the whole point of the control.

What happens when something breaks

Two questions an operations owner should ask before any segmentation product goes in front of a running process.

What happens if the enforcement point fails? With cloaking active, BlastShield fails closed and protected assets stop answering. That is a security property rather than a design flaw, and the availability answer is redundancy: deploy redundant Gateways. Already-established sessions are retained through failover, so live sessions do not drop.

What happens if the management plane is unreachable? The Orchestrator is the policy plane, not the data path. If it is unreachable the Gateway keeps enforcing the policy it currently holds, and the Orchestrator itself can be made redundant on-site, in the cloud, or both.

One honest limit, because an assessor will find it: the Orchestrator has no formal policy approval workflow. If a user’s permissions allow a policy change, that change is active immediately. Control is by permission model plus an after-the-fact configuration change log, not by two-person review. Separation of duties cannot be enforced within BlastShield. It has to be enforced through your own change management procedures.

How BlastWave segments OT networks

BlastShield delivers segmentation as a software-defined overlay. It enforces least-privilege microsegmentation without ACLs, VLAN re-architecture, or new hardware, in front of devices you cannot patch or agent. Any connection attempt from a system not explicitly configured with a policy is blocked and, where connection logging is enabled, logged, which is what bounds the consequence of an incomplete inventory. BlastShield does not perform asset discovery and does not claim to: it records what attempts to communicate, it does not enumerate what sits silent.

On hosts running the BlastShield Agent, BlastShield 2.0 adds self-isolation. With it on, the host accepts inbound traffic only over BlastShield, so other machines on the network cannot connect to it except through BlastShield.

Alongside segmentation, the same policy covers network cloaking, passwordless MFA, and secure remote access that replaces VPN with no inbound open ports and no concentrator, on a zero trust model.

The plain version: your PLCs stop answering scans.

BlastWave publishes compliance mappings for IEC 62443, NERC CIP, TSA Security Directives, CMMC Level 2, NIS2, the EU Cyber Resilience Act, and DoW OT Zero Trust. A mapping is not a certification, and we say so on every one of them.

Related: segmenting flat networks.

Hackopedia

Hackopedia

23 real OT cyber attacks. One recurring failure.

Explore Now

What is OT network segmentation?

OT network segmentation divides an operational technology network into isolated zones grouped by function and risk, and controls traffic between those zones with explicit policy, so a compromise in one zone cannot spread to the rest of the environment.

How is OT segmentation different from IT segmentation?

OT devices are frequently unpatchable, usually cannot run agents, and control physical processes that cannot tolerate downtime. Segmentation has to protect devices it cannot modify and cannot take offline, which rules out much of the re-addressing and agent-based tooling used in IT.

Does IEC 62443 require network segmentation?

IEC 62443 is built around it. The standard defines zones, groups of assets with a shared security requirement, and conduits, the controlled paths between them, and assigns security levels so each zone receives the protection it actually needs.

Can I segment an OT network without downtime?

Yes. An overlay that enforces policy in front of existing devices adds default-deny segmentation without re-IPing the network or taking the process offline, which avoids the outage that VLAN and firewall re-architecture usually requires.

Can you microsegment devices that cannot run an agent?

Yes. Agent-based microsegmentation cannot enforce between two devices that cannot host an agent, and a PLC cannot host one. BlastShield supports a Layer 2 isolation mode in which the switch forwards traffic to BlastShield, which enforces policy across Layer 2 switched segments in the same broadcast domain.

How long does a policy change take to take effect?

Seconds, and often under one second, from activation to enforcement. A comparable firewall or VLAN change moving through OT change control typically takes weeks.

Bring your own scenario. We will run a scan against a protected controller in front of you, then have an authorized operator reach the same device.

Request a demo