Using port check to inventory IoT and OT device exposure
AdminKit's port check tool tests nine commonly misconfigured TCP ports on the public IP you're connecting from, and while it can't enumerate every device at once, running it from each site you operate and matching results against a port-to-device table turns a one-off exposure check into a running inventory of what's reachable from the internet, and whether that has changed.
By AdminKit team, Netbase JSC · Updated
Key takeaways
- AdminKit’s port check tests nine TCP ports on the public IP the request comes from. One run describes one address at one moment, so an inventory is a series of dated runs, one per site or WAN connection.
- Each open port points to a device category (502 to a PLC, 554 to a camera, 7547 to an ISP-managed router). Your router’s port-forward and UPnP tables tell you which device it actually is.
- Keep a small record per run: address, date, open ports, inferred device, action taken. Re-run on a schedule and after any network change.
- A clean result covers those nine TCP ports only. UDP services and other TCP ports need a separate manual check, described below.
What is an exposure inventory built on port check?
An asset inventory for IoT and OT exposure usually means knowing, for every site or network segment you’re responsible for, what’s reachable from the internet right now — not a single snapshot of one address. AdminKit’s port check tool tests nine specific TCP ports against the one public IP a request comes from, by design, and never a target you type in, which means building an inventory with it is a matter of repetition and record-keeping rather than one all-at-once scan: run it from each site, connection, or public IP you operate, and keep each result as a dated entry rather than treating a single check as a permanent fact. This fits environments like a chain of retail locations, a factory with more than one WAN connection, or an integrator managing several customer sites — anywhere “exposure” is really a question about several addresses, not one.
Turning an open port into a device fingerprint
Each of the nine ports this check covers maps to a recognizable device category, which is the basis of a usable fingerprint table rather than a flat list of yes/no results:
| Port | Protocol | Typical device | Guide |
|---|---|---|---|
| 502 | Modbus TCP | PLC, RTU, or other industrial controller | Modbus TCP 502 remediation playbook |
| 1883 / 8883 | MQTT / MQTTS | IoT message broker | is port 1883 (MQTT) open? |
| 554 | RTSP | IP camera or video encoder | is port 554 (RTSP) exposed? |
| 23 | Telnet | Router, camera, or NAS admin shell | is port 23 (Telnet) exposed? |
| 7547 | TR-069/CWMP | ISP-managed router or ONT | TR-069 port 7547 exposure, explained |
| 80 / 443 / 8080 | HTTP(S) admin UI | Router, NVR, or other device with a web console | is port 8080 (admin panel) exposed? |
IANA’s port number registry is the authoritative reference for which protocol a given well-known or registered port is assigned to, which is useful when an open port doesn’t match the usual device in your environment — a non-standard service squatting on a well-known port is itself worth investigating. Two protocols common in building and sensor networks, BACnet/IP (47808) and CoAP (5683), are UDP-based and not covered by this check; both have their own guides — BACnet/IP exposure and is port 5683 (CoAP) exposed? — with a manual command to check them in the meantime.
How do I run the inventory, site by site?
The method is the same whether you look after two offices or forty kiosks. What changes is how many rows you collect.
- List the public addresses you are responsible for. One per site, plus any second WAN link, LTE failover router or separately contracted line. If you are not sure which address a site uses today, run what is my IP from a machine at that site.
- Run the port check from inside each site. The tool only ever tests the address the request comes from, so the check has to be started from a browser or script on that site’s network. A remote colleague, a site laptop or a jump host at the site all work.
- Record the result immediately, using the fields in the next section, before moving to the next site.
- Trace every open port to a device (see below), then decide: close, move behind a VPN, or keep with a written reason.
- Re-run after the fix to confirm the port now shows closed, and record that as a second dated row rather than editing the first.
Step 5 matters more than it looks. A record that shows “open on 4 March, closed on 6 March after removing the forward” is evidence that the exposure was found and handled. A record that was quietly overwritten shows nothing.
Building the inventory record
A minimal, workable inventory entry needs only a handful of fields per site or address: the public IP checked, the date, which of the nine ports came back open, the device type inferred from the fingerprint table above, and what action was taken (closed, moved behind a VPN, or deliberately left open with a documented reason). CISA’s Known Exploited Vulnerabilities Catalog is worth cross-referencing for any device type you find exposed, since several of the ports this check covers correspond to device families with a documented history of active exploitation once reachable from the internet.
A flat file is enough for a small estate. The rows below are a synthetic example using documentation addresses:
date,site,public_ip,open_ports,inferred_device,action,next_check
2026-03-04,warehouse-1,203.0.113.24,"502,8080",PLC + HMI web UI,forward removed; VPN only,2026-04-04
2026-03-06,warehouse-1,203.0.113.24,none,-,verified closed,2026-04-04
2026-03-04,shop-2,198.51.100.7,554,NVR camera stream,kept: vendor support; IP allowlist,2026-04-04
2026-03-04,shop-3,192.0.2.81,none,-,-,2026-04-04
Frameworks describe the same discipline in more general terms. The NIST Cybersecurity Framework 2.0 lists “inventories of hardware managed by the organization are maintained” as outcome ID.AM-01, and “representations of the organization’s authorized network communication and internal and external network data flows are maintained” as ID.AM-03. An exposure log like the one above is a narrow but concrete piece of ID.AM-03: it records which external flows into each site are possible, and whether each one is authorised.
How do I trace an open port back to a device?
The port check sees your public address, so an open port means your router is passing that port to something inside. There are three usual ways that happens, and each one leaves a trace you can read on the router:
| How the port got opened | Where to look | What you will find |
|---|---|---|
| A manual port-forward rule | Router or firewall NAT / port-forwarding page | External port, internal IP and internal port, often with a label someone typed years ago |
| UPnP or NAT-PMP, requested by the device itself | Router UPnP status page, or upnpc -l from a LAN host | A mapping created automatically, with the requesting device’s internal IP |
| The service runs on the router itself | Router admin settings (remote management, TR-069, VPN) | No forward at all: the router answers the port on its own WAN address |
Once you have an internal IP, match it to a device using your DHCP lease table, which normally shows a hostname and a MAC address. The MAC address prefix identifies the network hardware vendor, which narrows a mystery address to “a camera from vendor X” quickly; AdminKit’s MAC vendor lookup does that step. The third row is easy to miss: ports such as 7547 or 8080 are often answered by the router’s own management interface, so there is no forward to delete, only a setting to turn off or a call to make to the ISP.
On networks where UPnP is enabled, expect the open-port picture to change without anyone touching the router. Devices such as cameras and game consoles can ask the router to open a port for themselves. If your inventory shows a port appearing between two runs with no recorded change, the UPnP mapping list is the first place to check, and turning UPnP off on business sites removes that whole class of surprise.
What changes should trigger a re-check?
A monthly run catches slow drift. These events justify an extra run straight away, because each one can open or move a port without anyone intending it:
| Event | Why exposure can change |
|---|---|
| Router or firewall replaced, reset or reflashed | Default settings may re-enable remote management or UPnP |
| New device installed (camera, controller, gateway) | It may request a UPnP mapping or need a forward for remote support |
| Vendor or integrator remote session | Temporary forwards are often left behind after the job |
| ISP change or line upgrade | New CPE may expose TR-069 on 7547 or a web UI on 8080 |
| Public IP changed | The old row no longer describes the site; start a new one |
How do I check ports the tool does not cover?
The nine ports are the ones that most often turn up misconfigured on IoT and small-office equipment, not a full list. For anything else, run a manual check from a host outside the site, against that site’s own address only. For example, to look at common remote-access and database ports on TCP and a few UDP services:
# TCP: SSH, RDP, VNC, Redis, MySQL on your own public IP
nmap -Pn -p 22,3389,5900,6379,3306 203.0.113.24
# UDP: NTP, SNMP, CoAP, BACnet/IP (slower; needs root)
sudo nmap -Pn -sU -p 123,161,5683,47808 203.0.113.24
UDP results are less clear-cut than TCP ones, because a silent port can be either filtered or open and waiting for a protocol-specific request. The guides for port 123 (NTP), BACnet/IP and CoAP each give a protocol-aware probe that answers the question properly. SIP phones and PBXs also use UDP, and is port 5060 (SIP) exposed? gives an OPTIONS request that tests port 5060 the same way. Add these results to the same log, with a column for the method used, so a later reader knows which rows came from the tool and which came from a manual scan.
What can’t this process tell you?
Port check reports whether a TCP connection succeeded, not which internal device answers it — that mapping lives in your own router or firewall’s port-forwarding configuration, which only you have access to. It also can’t discover devices that aren’t port-forwarded to the internet at all, which is most of a typical IoT/OT network by design; this inventory specifically tracks internet-facing exposure, not a full asset list of everything on the LAN. Treat a clean result on all nine ports as confirmation that this specific exposure surface is closed today, not as a statement about the network’s overall security posture.
Next step
Run AdminKit’s open ports on my IP check from each site or connection in your inventory, starting with what is my IP to confirm the exact address being tested, and log each result against the fingerprint table above before moving to the next location. This inventory method is part of AdminKit’s machine protocols cluster of device and protocol exposure guides.
If you are a small business starting from scratch, the exposed devices checklist covers what to check first, including RDP and SMB, which port check does not test.
This guide is general information, not a security audit of your specific network — see our disclaimer for what AdminKit's guidance does and does not commit to.
Frequently asked questions
Can port check scan every IoT device on my network in one pass?
No. It scans a fixed list of nine TCP ports against exactly one address: the public IP the request is coming from, determined server-side, with no field to type in another target. It can't enumerate internal device addresses behind your router, and AdminKit's Check API itself refuses a ports scan whose target isn't the caller's own IP. Building an inventory with it means running it once per public IP you operate (per site, per WAN connection) and treating each result as that location's current exposure snapshot, not a single scan of your whole fleet.
If port check shows one port open, how do I know which specific device is behind it?
The check tells you a port is open on your public address, not which internal device a router's port-forward rule sends that traffic to — only your own router or firewall's forwarding table can tell you that. What the open port does tell you, combined with which protocol commonly uses it, is the type of device worth looking for: port 502 open points you toward a Modbus-speaking PLC or RTU, port 1883 toward an MQTT broker, and so on. Cross-check against your router's port-forwarding rules to find the specific internal device.
Does this check cover every protocol an IoT or OT device might expose?
No — it currently probes nine TCP ports only: 23 (Telnet), 554 (RTSP), 1883/8883 (MQTT/MQTTS), 502 (Modbus), 7547 (TR-069), and 80/443/8080 (admin web UIs). BACnet/IP (47808) and CoAP (5683) are UDP-based and not covered by this tool — AdminKit's dedicated guides for those two ports include a manual nc or UDP probe command to check them separately.
How often should I re-run this for an inventory to stay useful?
There's no fixed interval that fits every environment, but an inventory is only as good as its last check: firmware updates, a new port-forward added for a one-off maintenance job, a swapped router, or a new device joining the network can all change what's exposed without anyone updating a spreadsheet. Re-checking on a fixed cadence (monthly is a reasonable default for most small sites) and after any known network change catches drift that a single, never-repeated scan can't.
My public IP changed since the last check — does my inventory still apply?
Not necessarily to the same address. A dynamic IP or one behind Carrier-Grade NAT can change or be shared with other customers between checks, so a result tied to an old address may no longer describe your current exposure at all — see what is CGNAT for how to tell whether your address is shared, and confirm your current public IP with AdminKit's my IP tool before treating an older inventory entry as still accurate.
Want AdminKit to watch this for you?
Free checks tell you the state today. Protective DNS and monitoring tell you when it changes.
Talk to an expert
Huy Nguyen (David), Founder & CEO, Netbase JSC
Ask whether this fix applies to your setup, or get a second pair of eyes on your DNS and mail records.