top of page

Converged IT/OT Security with Zero-Trust Architecture

Dec 1, 2024
7 min read

Updated: Oct 12

IT and OT convergence illustration with interlocking chain, information technology and operational technology icons

An engineer needs to review an HMI alarm from a managed workstation. The workstation is on the company network, but that location alone does not answer whether the engineer should reach this machine, which commands are allowed, or how the access will be removed after the job.

That is the problem zero-trust architecture addresses. It makes access to a resource depend on an explicit decision about the user, device, request, and policy. In an operational technology (OT) environment, that decision must also preserve the availability, reliability, and safety needs of the process.

Key Takeaways

  • Zero trust does not assume that a user or device is trustworthy because it is inside a corporate or factory network. Authentication and authorization are checked for access to a specific resource [1].

  • IT/OT convergence does not require connecting every controller to an identity service or cloud platform. Use managed access boundaries and keep local control functions within their approved design.

  • A router, firewall, VPN, jump host, or access service can be one enforcement point. No single appliance proves that an organization has a complete zero-trust architecture.

  • Network segmentation remains useful, but a VLAN or firewall rule alone does not make an access decision about a particular person, device, or resource.

  • Start with an inventory of users, assets, flows, owners, and operational limits. Stage and test policy changes before expanding them across production systems.

What Zero Trust Changes

NIST describes zero trust as an approach that does not grant implicit trust to a user or asset solely because of its physical or network location or because the organization owns it. Authentication and authorization of both the subject and device occur before a session to an enterprise resource is established [1]. The focus is the resource and the access decision, rather than assuming that all traffic inside a network segment is safe.

This does not mean every industrial controller must run a modern identity agent, or that networks should be connected without boundaries. Many operational devices have fixed functions, long service lives, and strict change windows. NIST's OT guidance calls out the distinct performance, reliability, and safety requirements that security work must account for [2].

For an OT deployment, apply the access principle at a boundary the site can operate. The exact placement depends on the asset, the control design, and the organization's identity and monitoring services. This is a design decision, not a feature implied by the words “zero trust” on a product page.

Capabilities to Verify at the IT/OT Boundary

The five capability areas below are deployment roles to evaluate, not features guaranteed in every industrial router. Identify the responsible component, its supported scope and its operational owner before relying on a product label.

Capability area

Role in the deployment

What to verify

Policy decision and enforcement

Identity and policy services decide access; an appropriate gateway or host enforces it.

Identify decision, enforcement and revocation components, their inputs and logs; a router alone is not the whole architecture.

Protocol-aware segmentation

Restrict approved industrial flows and, where supported, specific operations.

Verify exact protocol versions, inspection visibility, encryption limits and permitted functions. Do not assume every router inspects Modbus, PROFINET, DNP3 or OPC UA commands.

Threat monitoring and analytics

Collect relevant logs and assess activity against the approved flows, using suitable monitoring components.

Verify telemetry, detection scope, integration and operating resources. Embedded machine learning is a product-specific capability, not a general router property.

Controlled incident response

Revoke access or contain an affected path through an authorized operational process.

Validate process and safety effects, approval ownership, logging and rollback before automated changes; do not assume SOAR integration or safe automatic PLC isolation.

Resource-specific remote access

Give a named user and managed device only the access required for an approved resource and task.

Verify authentication, authorization, access duration, session logging and removal. Use a supported boundary or engineering host where an OT endpoint cannot run an identity agent.

Before automated containment changes an OT communication path, validate the effects on process availability and safety, name the responsible owner and test the rollback.

Draw the Access Path and Its Owners

Before buying a security product or changing routes, draw the path for one real task. For example:

named engineer on managed device → identity and access policy → monitored enforcement point → OT access zone → approved engineering workstation or target

For the concrete maintenance task, apply the PLC and HMI remote-access workflow and test the engineer-to-device application path.

For each step, record the user or service identity, device state required for access, target resource, allowed protocol, approval owner, access duration, and logs that operations can review. Identify which system makes the decision, which system enforces it, and who can revoke the request.

The path may include an identity provider, remote-access service, VPN gateway, firewall, OT DMZ, jump host, and engineering workstation. They are separate roles even if a deployment combines several functions. NIST's zero-trust model names a policy engine and policy administrator as decision components and a policy enforcement point as the component that enables, monitors, and terminates the connection [1].

Keep the scope narrow. If the task is to read an alarm, do not grant a general route to every PLC and HMI. Separate engineering access from ordinary telemetry and from any control path that can change process behavior. Record what happens when the identity service, access gateway, or WAN is unavailable; local control should not depend on an untested remote-access service.

Use Segmentation Without Treating It as Trust

Zones, firewalls, VLANs, and access-control lists can limit which systems communicate. They help create boundaries between business IT, OT supervisory systems, and control networks.

They do not, by themselves, verify the person or managed device requesting a session, nor do they establish that the requested operation is appropriate.

For network implementation, design VLANs and permitted inter-VLAN traffic around the required sources, destinations and services.

Use network controls to enforce the paths defined by the access policy. Permit only the required source, destination, protocol, and direction at each boundary. Keep configuration and logs so that the site can explain why a path exists and who approved it.

If a device cannot support individual authentication, the surrounding access service or controlled engineering host may provide a place to apply and audit the decision. Confirm that this mediation fits the equipment vendor's operating requirements.

Avoid broad “trusted OT subnet” exceptions that give any connected workstation reach to every controller. Also avoid changing the control network without an approved maintenance window, a tested backup, and a rollback owner. OT security guidance treats architecture, boundary protection, and operational constraints together [2].

Roll Out Access Policies in Stages

Use a sequence that operations can review and reverse:

  1. Inventory assets and access. List the controllers, HMIs, engineering stations, gateways, services, users, vendors, and current flows. Note which assets cannot be changed or monitored directly.

  2. Map the process boundary. Identify the systems that must remain local for control and safety, the supervisory systems that consume their data, and the approved paths between zones.

  3. Assign decision and enforcement owners. Name who verifies user and device identity, who approves resource access, which gateway or host enforces the decision, and who reviews logs and revocations.

  4. Write the minimum policy. Start with a named user or service, one resource, the necessary protocol, and a defined access period. Document the restrictions and connection requirements for each remote-access type, and authorize it before enabling the connection [3]. Record any exception and its owner.

  5. Test in a non-production path or maintenance window. Verify allowed access, blocked access, log visibility, local operating behavior, and the rollback steps. Include representative device failure and access-service failure cases.

  6. Expand only after review. Compare the observed flows with the approved design. Remove unused paths and credentials, then repeat tests after material changes to identity, gateways, firewall rules, firmware, or controller networks.

The work is not complete when a dashboard shows fewer network routes. Acceptance depends on whether authorized operators can perform approved work, unauthorized paths are blocked, local control remains within its design, and the people responsible can detect and recover from a failure.

Keep Related Network Decisions Separate

Zero trust does not choose the site's VPN protocol, carrier, or WAN recovery method. Those decisions need their own endpoint, routing, and failure tests. If machine networks reuse the same IP ranges, resolve that addressing issue before granting cross-site access; a policy engine cannot make overlapping routes unambiguous. See How to Connect Multiple PLCs with the Same IP Using NAT and VPN for the addressing design.

If access depends on multiple WAN links, test the remote session and the site's recovery behavior when a path fails. The expected result belongs to the application and operations owners, not a general claim that redundancy is seamless. See Failover vs. Load Balancing vs. Bonding for 5G and Starlink for the separate multi-WAN decision.

Conclusion: Make Each Access Decision Explicit

Zero-trust architecture is a way to decide access to resources; it is not a router feature label or a replacement for OT network design. Start with one user, one device, one operational task, and one target. Place the enforcement point where the site can maintain it, limit the permitted path, and validate both access and denial without disrupting local control.

Before expanding a policy to more production equipment, have the controls, security, and operations owners review the asset inventory, allowed flows, failure cases, and rollback plan together.

FAQ

Does zero trust mean trusting nothing on the network?

It means network location or ownership alone does not grant implicit trust. NIST's model requires authentication and authorization of the user or other subject and the device before a resource session is established [1]. The policy still needs to define which requests are allowed.

Does zero trust replace a VPN or firewall?

No. A VPN can protect a communication path, and a firewall can enforce network boundaries. They may be components of a wider design, but they do not automatically provide every identity, device, resource, logging, and lifecycle decision required by that design.

Is a VLAN or OT DMZ the same as zero trust?

No. Segmentation can restrict reachability between systems. It does not by itself authenticate a person and device for a specific resource request. Use segmentation as an enforcement layer alongside access policies and operating procedures.

Can an industrial router provide a complete zero-trust architecture?

Do not assume so. A router or firewall may enforce part of a path if its exact model, firmware, identity integration, policy scope, and logs meet the requirement. A full architecture also depends on identity services, resource ownership, access lifecycle, monitoring, and OT operations.

Should every PLC or HMI run zero-trust software?

Not necessarily. Some operational devices cannot host modern identity agents or accept frequent updates. Define a controlled access boundary around those resources and verify that the intermediary fits the equipment and process requirements.

How should a factory begin a zero-trust project?

Start with an asset and flow inventory, identify one remote task that needs tighter access, name its owner, and design a narrow policy and enforcement path. Test the allowed task, denied paths, logging, service failure, and rollback before expanding the scope.

References

Recent Posts

See All
bottom of page