Pricing Check your website Blog
15 November 2026

CVE-2026-129444 - hard-coded FTP credentials on a 2013 D-Link gateway

Two records reached CVSS 10.0 in the twelve hours this pick was drawn from, and this is the one whose vector claims the most: a hard-coded-credential flaw in the FTP daemon of the D-Link DSL-6740C, scored 10.0 by the assigning CNA under CVSS 4.0, remotely reachable with no privileges and no user interaction. A proof-of-concept is public. Whether any of these devices are actually reachable from the internet is the question this post cannot answer, and that is worth more attention than the score.

What the bug is

The record locates the flaw in /etc/templates/ftpd/ftpd_conf.php, part of the FTP Daemon component of firmware 20131126. The published description is thin and says so: it affects an unknown function of that file, and manipulating it leads to hard-coded credentials. The classification is CWE-798 (hard-coded credentials), with CWE-259, use of a hard-coded password, as a second category.

What makes it a 10.0 is the vector, not the description:

  • AV:N, network. Reachable over the network, not from an adjacent network or locally.
  • AC:L, low complexity. No special conditions on the target have to hold.
  • AT:N, no attack requirements. Nothing has to be pre-arranged on the device before the attack works.
  • PR:N, no privileges. The attacker authenticates as nobody.
  • UI:N, no user interaction. Nobody has to click, log in or upload anything.
  • VC/VI/VA all High, and SC/SI/SA all High: the claim is that impact is total on the vulnerable component and spills into the rest of the device.
  • E:P, proof-of-concept. The exploit exists as published code, in a public evidence repository, not as observed exploitation.

That combination is what a 10.0 under CVSS 4.0 means: the metric set leaves nothing between the internet and full impact. It is a claim about the worst case, made by one identity.

Who it affects

D-Link DSL-6740C, firmware 20131126, with no platform list and no other products in the affected entry. The version is a build date: 26 November 2013. The record lists no unaffected version, so there is no fixed release to aim at, and a device built to twelve-year-old firmware is unlikely to be collecting security updates even if the vendor intends otherwise.

For a CPE-less, single-build entry like this, the practical scope question is narrower than "which firmware": it is whether the FTP service is switched on. That is a configuration fact about your own estate, not a fact in the record.

What is exposed

Unmeasured. The exposure source configured here is a Censys Platform account on the Free tier, and the Platform API answers search and aggregate calls with 403: This endpoint requires an organization ID for API access. Free users can only access this endpoint through the Platform UI. Lookup by IP works; internet-wide counting does not. So this post states the absence rather than printing a reassuring zero.

The query that would size it, once run against an entitled account, is a device or banner match on port 21 rather than a CVE match: no scanner knows this CVE yet, it is twelve hours old. Counting hosts whose FTP banner carries this device family answers the population question directly, and the result should be read with the usual caveat that a scanner only counts what it happened to see on the day.

For readers who run these devices, the honest substitute for a global count is a local one: does your asset inventory contain a DSL-6740C, and is TCP 21 open on it.

Where the 10.0 comes from

The cache holds six score rows for this record. Four of them are the same claim re-expressed under different CVSS versions, and one of them is not 10.0:

Score Version Identity Feed
10.0 4.0 VulDB, the assigning CNA cvelistV5 CNA container
10.0 3.1 VulDB, the assigning CNA cvelistV5 CNA container
10.0 3.0 VulDB, the assigning CNA cvelistV5 CNA container
10.0 3.1 [email protected] NVD 2.0 feed
10.0 2.0 [email protected] NVD 2.0 feed
9.3 4.0 [email protected] NVD 2.0 feed

The last row is the same claim carrying the full modifier and threat block - E:P plus every MAV, MVC and Security Requirements term written out as X - and it scores 9.3 instead of 10.0. Nothing about the flaw differs between those rows; what differs is which rating of the CVSS 4.0 family is being reported. A filter that compares CVSS numbers across feeds without comparing the vector text will read them as two different opinions about severity.

Worth noting what is absent: no NVD-generated score of its own, and no CISA-ADP row. The 10.0 is the CNA's, carried by NVD unchanged. CISA's KEV catalogue holds 1,781 entries as of 2026-11-13 and does not include this CVE, which is unsurprising for a record this young. See who actually scores a critical CVE for how often that pattern holds.

What to do

Disable the FTP service on these devices unless something concretely needs it. That removes the entry point entirely, and for a twelve-year-old firmware build it is the only mitigation that does not depend on a vendor patch that the record does not name.

If FTP has to stay, restrict TCP 21 at the perimeter to the addresses that need it, and treat the device as unpatchable: assume the credentials in the shipped configuration are known to anyone who reads the firmware image. If you cannot rule out a DSL-6740C in your estate and cannot reach it from your own scanning range, that is the finding, not the score.

Want this depth applied to your own systems?

Automated security testing from €100 with a fixed price agreed before testing begins, or see what your own website gives away in a single request.

← All posts