Your External Port Scan Can’t Prove Your Network Is Safe

Date:

You pointed a port scanner at your home’s public address from the outside, and it came back clean: everything closed or filtered. You concluded that nothing at home is reachable from the internet. That conclusion, not the scan, is the mistake.

The short answer: an external port scan only tests the ports it was told to test, so a clean result is the absence of a hit on a sampled list, not proof of no exposure. Three things it routinely misses: a port forward to an uncommon high port that wasn’t in the scanned range, a port that a device on your LAN opened for itself through UPnP without you knowing, and your ISP’s remote-management channel, which is reachable even though its inbound port is filtered because it is dialed outbound. The reliable check is not a better scan. It is to read your router’s own forwarding and UPnP tables directly, and confirm each entry is one you meant to create.

Checking exposure by scanning from outside is appealing because it feels like the attacker’s view. And it is useful, as far as it goes. But a scan answers a narrower question than the one you are asking. You want to know “is anything at home reachable from the internet?” A scan answers “did any of these specific ports respond when I probed them just now?” The gap between those two questions is where a real exposure hides, quietly, behind a green result.

Step 1: Know what your scan actually tested

Every scanner tests a list of ports, and the default list is smaller than people assume. By default, nmap scans the 1,000 most common ports, not all 65,535. The online “is my port open” checkers test far fewer, often a dozen well-known ones. So a clean result from a default scan means “none of the common ports answered,” which is a long way from “nothing is listening.”

# Default: tests only the 1,000 most common ports
nmap 203.0.113.10

# Full: all 65,535 ports - slower, but nothing hides in the high range
nmap -p- 203.0.113.10
Code language: Bash (bash)

A clean scan means the ports it tested were quiet. It says nothing about the ports it didn’t test.

This is the same shape as a success message that isn’t confirmation: the reassuring signal reports only on the narrow thing it checked, and you read it as a verdict on the whole system. The fix in both cases is to stop trusting the summary and look at the authoritative source.

Three exposures an external port scan won't show youA clean scan is the absence of a hit on a sampled list, not proof of no exposure
1. Forward to a high port

A rule sends inbound 31400 to a device. A common-ports scan never tests it.

2. A port UPnP opened

A LAN device asked the router to open a WAN port for itself – no approval, no dialog, no record you see.

3. An outbound channel

TR-069 dials OUT to the ISP and takes config. Inbound port reads closed; the channel is wide open.

The reliable check is the router's own forwarding, UPnP and remote-management tables – not a scan. A scan is a lead; the configuration is the answer.

Step 2: The forward it couldn’t see

Here is a concrete miss. Suppose a device on your network, a small single-board computer running a peer-to-peer service, needs inbound connections, so at some point a forwarding rule was added on the router: WAN traffic on port 31400 goes to that device. Port 31400 is nowhere near the common-ports list. A default scan sails right past it and reports clean. The forward is live the entire time.

# Default scan: the forward on 31400 is nowhere near the common ports
$ nmap 203.0.113.10
All 1000 scanned ports are filtered          # looks clean

# Full scan: now it shows
$ nmap -p- 203.0.113.10
PORT      STATE SERVICE
31400/tcp open  unknown                      # there it is - open the whole timeCode language: PHP (php)

Worse, that rule often points at an IP, not a device. If the forward targets 192.168.1.3 and that address sits inside the router’s DHCP pool rather than being reserved, then the rule follows the number. The day your DHCP hands 192.168.1.3 to a different device, the forward now exposes that device’s port 31400 to the internet, and nothing told you. A scan on the day you check might miss the port; even if it hits, it can’t tell you the rule is aimed at a moving target.

Step 3: The port UPnP opened for itself

UPnP lets a device on your LAN ask the router to open a WAN port on its behalf, with no human approval. A game, a console, a torrent client, or a compromised IoT gadget can open a hole and you will never see a dialog about it. If your external scan happens to hit that port, you’ll find an open port with no idea why it’s there; if the scan’s list misses it, you won’t find it at all.

So “my scan was clean” and “no device opened a port behind my back” are two separate claims, and the scan only ever speaks to the first. The way to settle the second is to read the router’s UPnP mapping table and confirm it is empty, or holds only mappings you understand. You can list the active mappings from any machine on the LAN with the miniupnpc client:

# List the WAN ports UPnP has opened on your router
upnpc -lCode language: Bash (bash)

Every row is a port some device asked the router to open. Many people simply turn UPnP off, which removes the whole class of silent, self-service holes.

Step 4: The channel that answers outbound, so a scan can’t find it

Your scan tests inbound reachability: it sends packets at your public IP and watches for a reply. But some exposure doesn’t wait to be connected to. Many consumer routers run TR-069 / CWMP, an ISP remote-management protocol. The router dials out to the ISP’s management server on a schedule and accepts configuration pushes over that connection. Its inbound port (commonly 7547) is usually filtered, so your scan sees it closed and you feel safe. The closed port is irrelevant: the management session is outbound-initiated, so the ISP can reconfigure the router regardless of any inbound firewall.

That isn’t necessarily something to panic about, but it is something a scan can never reveal, and it means “inbound 7547 is filtered” and “my ISP can’t reach this router” are, again, two different statements. Only one of them is true. It is the same trap as a missing page that still returns 200: a status that reads healthy while the reality it is supposed to describe is anything but.

Step 5: Check the source, don’t infer it

Reading the router’s own configuration takes a minute and answers the question the scan only gestured at. Open the admin interface and look at four things:

  • Port forwarding / virtual servers. The table should be empty, or every rule should point at a reserved IP you recognize, on a port you meant to open. A rule aimed at a DHCP-pool address is a latent exposure even if it looks harmless today.
  • DMZ. It should be off. A DMZ host forwards everything inbound to one machine and undoes every other precaution in one setting.
  • UPnP. Off, or a mapping table you can account for line by line.
  • Remote / WAN management. Off for the web admin. Note that TR-069, if present, is usually ISP-locked and outbound, so judge it separately: you can’t disable it, but you should know it’s there.

If you still want an external scan as a second opinion, run a full one (all 65,535 ports, not the common list) from a host outside your network, and treat a hit as a lead to chase in the router’s tables, never a clean result as a clearance. The scan is a smoke alarm, not a safety inspection.

The short version

An external port scan tests a sample of ports at one moment, so a clean result proves only that the sampled ports were quiet, not that nothing is exposed. It routinely misses a forward to an uncommon high port, a port UPnP opened without your knowledge, and an ISP management channel that answers outbound while its inbound port reads closed. Don’t infer safety from the scan. Read the router’s own forwarding, DMZ, UPnP, and remote-management settings, and confirm every entry is one you meant to create. A scan is a lead, the configuration is the answer.

Frequently Asked Questions

Is an external port scan useless, then?

No. It’s a useful second opinion, and a hit is genuinely informative: something answered on that port, and you should go find out what. The error is only in the other direction, reading a clean result as proof of no exposure. Treat a scan’s green as “nothing obvious on the ports I tried,” not as a clearance, and run it over the full port range rather than the common-ports default if you’re going to rely on it at all.

If I just scan all 65,535 ports, am I covered?

A full scan closes the “uncommon high port” blind spot, which is the biggest one, but not the other two. It still can’t tell you why a port is open (so a UPnP-opened hole looks the same as a deliberate one), and it still can’t see an outbound-initiated channel like TR-069 whose inbound port is filtered. A full scan plus a read of the router’s forwarding and UPnP tables together answer the question; the scan alone does not.

How would a port end up open without me opening it?

Most often UPnP: a device on your LAN asked the router to open a WAN port for itself, silently, which is exactly what UPnP is designed to allow. Less often, a forwarding rule you created long ago for a device that has since changed its address, so the rule now points somewhere you didn’t intend. Both are invisible to you until you read the router’s tables, which is why the configuration, not the scan, is the source of truth.

What’s the single most useful thing to check?

The port-forwarding table, and specifically whether each rule points at a reserved IP. A forward to an address inside the DHCP pool is the quiet trap: it works today, looks deliberate, and silently re-aims at whatever device next receives that address. Reserve the addresses your forwards depend on, or delete the forwards you no longer need.

Share post:

Subscribe

Popular

More like this
Related

Can AI CLI Conversation History Leak Answers Into a Blind Test?

You use an AI command-line tool to grade or answer hundreds of questions, one fresh run each. Then you find out it keeps every conversation on disk, answer keys included. Is run 200 quietly reading run 199?

The Code Snippets 403 That Comes From Your Host’s Firewall, Not WordPress

You change a few lines in a snippet, press Save, and get "Could not update snippet. Request failed with status code 403". Press Save on the unchanged snippet and it works. Your login is fine. Something between you and WordPress is reading the code itself.

Why a Flaky Test Passes Alone and Fails When the Machine Is Busy

Run the test by itself and it passes, every time. Run the whole suite in parallel and it fails, but only some days, and its failure looks exactly like a real bug. In our case the code was fine. The test was waiting on a clock.

Why Multisite Subsite Registration Sends Visitors to the Main Site

You put a "create an account" link on one site of a WordPress multisite. Visitors who click it land on the main site, which either says "Registration has been disabled" or signs them up there instead.