Internet facing OT devices are rarely the ones anybody meant to put online. They are the 4G failover router a contractor installed for remote diagnostics, the building management web interface a facilities supplier needed access to, the vendor VPN gateway built for commissioning and never decommissioned. Almost every article on this subject reaches the same conclusion, which is that operational technology should not be reachable from the public internet. Very few explain how to find the devices nobody wrote down.
This blog outlines seven checks that surface exposed industrial kit from the outside, in the order an attacker would work through them, with the verification steps that keep the exercise safe on a live plant.
What Counts as an Internet Facing OT Device?
The formal definition is narrow: an operational technology asset with a routable IP address that can be reached from an untrusted network without first traversing a controlled boundary. In practice, exposure is messier than that. A human machine interface sitting behind a forwarded port on a site firewall is internet facing. So is a building management system login page published for a facilities contractor, a cellular gateway at a remote pumping station with a public SIM, and a jump box that exists only so an integrator can dial in on demand.
Air gaps rarely survive contact with a maintenance contract. They are broken by a vendor support clause requiring remote diagnostics, by a 4G backup link installed for resilience, or by a site refurbishment where the new network was commissioned quickly and the temporary rules were never reversed. The gap is almost never deliberate, which is precisely why it does not appear in the asset register.
Industrial protocols announce themselves when they reach the public internet. Modbus TCP listens on port 502, DNP3 on 20000, BACnet/IP on 47808, EtherNet/IP on 44818, and Siemens S7comm on 102. Alongside them sit the HTTP and HTTPS management interfaces on 80, 443 and 8080 that often matter more, because a web login page with default credentials is a far easier route in than a raw protocol port. The National Cyber Security Centre has repeatedly warned about increased interest in internet-exposed operational technology in the UK, and CISA advisories in the United States have made the same point about leaving programmable logic controllers directly reachable.
One distinction shapes everything that follows. A device you own, a device your integrator owns, and a device sitting on a landlord's shared network all demand different responses, different authority to act, and different legal footing before anyone scans anything.
Check 1: Build the Outside-In Asset Picture Before the Inside-Out One
Start from IP ranges, autonomous system numbers, registered domains and the geography of your sites. Do not start from the configuration management database, because unknown assets are missing from it by definition. An inventory built inside the perimeter describes what the organisation believes it operates. An inventory built from outside describes what an attacker can actually see, and the difference between the two is the real finding.
Ownership verification comes first, before any monitoring begins. On a shared industrial estate, a public IP adjacent to yours may belong to the unit next door. Kit installed by an integrator may sit on the integrator's own connectivity. Scanning something you do not own is both a legal problem and, on a control network, a safety problem. Confirming the range, the device and the domain belong to you is not administrative overhead, it is the step that makes the rest defensible. Structured external asset discovery and attack surface mapping is designed around exactly this sequence: find, attribute, then watch.
Unknown OT assets cluster in predictable places. Acquired sites inherited with their own connectivity and their own contractors. Decommissioned facilities that are still powered and still routed. Pilot plants and test lines built outside the main change process. Temporary works connections laid in for a project that quietly became permanent.
Check 2: Hunt the Remote Access Paths, Not Just the Devices
The exposed controller makes the headline, but the way in is usually an access path. Vendor VPN accounts created during commissioning and never revoked. Jump boxes built for a single upgrade weekend. Engineering workstations that bridge the IT and OT networks and happen to be running exposed RDP, VNC or a web-based remote desktop portal.
Cellular routers deserve particular attention. A SIM-connected gateway with a public address and a default management password is a straight line from the internet into the control network, and because it was often procured by the engineering budget rather than IT, it may never have been added to any register. Our guidance on closing exposed RDP monitoring blind spots applies directly here, because the same oversights that leave remote desktop open on a corporate estate leave it open on an engineering laptop.
So, when building the exposure picture, treat every remote access mechanism as an asset in its own right: who created it, who uses it, what it reaches, and whether it has multi-factor authentication in front of it.
Check 3: Separate Live Plant From Test Rigs, Honeypots and Ghosts
Not everything that looks like exposed OT is exposed OT. Research honeypots are deliberately seeded across the public internet to attract industrial scanning. Integrators run demonstration kit. Shared hosting on the same range produces coincidental matches. Reporting a honeypot to a plant manager as a critical finding costs credibility that is difficult to recover.
Proportionate assessment matters more in operational technology than anywhere else. A legacy controller can fall over under a standard web application scan, and a controller falling over is a production incident with safety implications. Scanning boundaries should be customer-controlled, with explicit exclusions, and some assets should be identified from banner and certificate evidence and then watched rather than actively probed. That is the principle behind managed infrastructure and web application scanning with defined scope, and it is set out further in our approach to responsible external assessment.
Evidence quality determines whether anything gets fixed. A plant manager asked to authorise a firewall change at short notice will want the IP, the port, the banner or certificate evidence, the date observed, the attribution reasoning and the specific process the device controls. Collect that before raising the request, not afterwards.
Check 4: Look for Your OT Estate in Leaked Credentials and Criminal Chatter
Infostealer logs are where industrial exposure becomes urgent. A stealer log harvested from an engineer's laptop may contain saved browser credentials, session cookies, a SCADA portal URL and the VPN profile used to reach it. Access brokers on criminal forums and Telegram channels frequently advertise industrial targets by sector and country rather than by company name, describing the access on offer rather than naming the victim, which means attribution has to be done by matching technical detail against your own estate.
The highest-priority finding is almost never a single exposed device. It is a leaked credential that matches a remote access portal you have just confirmed is reachable from the internet with no second factor. Read in separate reports, those two items look like a routine password reset and a low-severity exposure. Read together, they describe a working route into the control network. Credential intelligence and asset discovery belong in the same view, which is also why coverage gaps in leaked credential monitoring are worth closing before anything else.
Contractor and integrator identities are part of this picture. An engineer who uses the same credentials for their own company and for your plant portal extends your exposure to an organisation whose device estate you do not control. Separating internal identities from those issued to external parties limits the blast radius considerably.
Check 5: Track Drift, Because Exposure Comes Back
OT exposure recurs on a predictable cycle. Firewall rules are reopened during maintenance windows so a vendor can complete a job over a weekend. Commissioning works introduce temporary routing. A planned outage leads to a management interface being published for a few days, and the ticket to close it afterwards is written but not completed. Annual audits miss all of this, because the exposure opens and sometimes closes between audits.
Continuous change detection catches it instead: new ports appearing on known IPs, fresh certificates on management interfaces, DNS records pointing at site ranges, new addresses appearing where none existed last month. Continuous asset monitoring with change detection and alerting works best in OT when it is tuned low volume and high signal, with duplicates suppressed rather than forwarded.
Duplicate findings kill industrial remediation faster than anything else. One device appearing as five findings across two scanners teaches the plant team that the report is noise, and once that happens the next genuine finding is ignored as well. Reconciliation is best run monthly and aligned to the site works calendar, not the IT patch cycle, because that calendar predicts when new exposure is most likely to appear.
Check 6: Assess the Suppliers Who Hold Access to Your Plant
Ask three questions of every integrator, maintenance provider and managed OT vendor: who holds remote access to our sites, through which mechanism, and how is that access revoked when a person leaves. Many organisations discover that the answer is held informally by one site engineer rather than in any contract.
Public security signals support the conversation. Certification against recognised standards such as ISO 27001 or Cyber Essentials, historic breach disclosures, the state of the supplier's own external estate and whether their published services show obvious exposure all indicate how carefully they are likely to treat your access credentials. Our note on the signals worth checking in vendor security monitoring covers the practical version of this assessment.
When the exposed device turns out to belong to a supplier, the response changes but the obligation does not. The asset still touches your process, and under the regulatory direction set by the Cyber Security and Resilience Bill, which is progressing through Parliament at the time of writing, expectations on operators of essential services and their supply chains are tightening rather than loosening.
Check 7: Write the Response Path You Can Run at Four O'Clock on a Friday
Finding an exposed human machine interface on a Friday afternoon is a governance problem as much as a technical one. Somebody has to decide whether the device comes off the internet now or at the next planned outage, and that decision requires knowing what the process does without it.
A workable response path records four things in advance: who can authorise taking a device offline, who must be consulted on the engineering side, what the fallback is for the process the device controls, and what the interim mitigation looks like if removal has to wait. Interim options usually include restricting access to a small set of source addresses, placing the interface behind an authenticated gateway with multi-factor authentication, or rotating credentials and enabling logging while the permanent fix is scheduled.
Group related findings into one remediation project per site or per vendor rather than issuing dozens of individual tickets. Then report each finding twice: as a work order for the engineering team, with IPs, ports and the physical location, and as a risk statement for the board, describing the process at stake and the plausible consequence. The same finding, written for two audiences, moves far faster than one technical report sent to both.
What Good Looks Like After 90 Days
Three months of consistent effort should produce a reconciled external inventory per site, with a named owner for every public IP associated with the organisation. Remote access paths should be reduced to a known, logged and time-bound set, with vendor accounts enabled on request rather than standing open. Monitoring should flag new exposure within days of a site change, not at the next annual audit.
The organisations that get there treat discovery as continuous rather than as a project. Exposure is created by ordinary operational activity, and it will be created again next quarter by the next refurbishment. To see how your own external estate looks from an attacker's position, including the internet facing OT devices nobody put on the register, start with the DarkInvader external attack surface platform.
Frequently Asked Questions
What counts as an internet facing OT device?
Any operational technology asset reachable from an untrusted network without passing through a controlled boundary, including HMIs behind forwarded ports, BMS web interfaces, cellular gateways with public IPs and vendor remote access portals. Industrial protocol ports such as Modbus TCP on 502 or S7comm on 102 are the obvious indicators, but management interfaces on 80, 443 and 8080 are more commonly the exposed element.
How can I find internet facing OT devices I do not know about?
Begin from IP ranges, ASNs, domains and the geographic footprint of your sites rather than from the asset register, because unrecorded assets are absent from the register by definition. Reconcile what is visible externally against the OT inventory and investigate every difference, paying particular attention to acquired sites, remote stations and locations that have undergone recent works.
Is it safe to scan OT devices for exposure?
Standard scanning techniques can disrupt legacy controllers, so assessment of operational technology should be proportionate and scoped with customer-controlled boundaries. Confirm ownership before any activity, use passive banner and certificate evidence where a device may be fragile, and agree exclusions with the engineering team before scanning begins.
What should I do first if I find an exposed PLC or HMI?
Verify ownership and confirm the device is live plant rather than a honeypot or demonstration rig, then record the evidence you will need for a change request. Contact the site engineering lead before altering anything, because taking a controller offline without understanding the process it manages can create a safety incident rather than preventing one.
Who is responsible when the exposed OT device belongs to a supplier?
Accountability for the process remains with the operator even when the asset belongs to an integrator or landlord. Raise the finding with the supplier in writing, ask who holds remote access and how it is revoked, and use contract review to establish security obligations and notification requirements for future works.













