Skip to content

WordPress plugin · High, 7.2 of 10 · patched

The guest list that was only half checked

A database search-and-replace plugin with 100,000+ installs limits an administrator to the tables of their own site. In 2024 a fix added a guest list of allowed tables. We read that fix the way a detective reads an alibi, and the list did not check every name it was given. On a WordPress network, the administrator of one small subsite could make themselves Super Admin of all of it. It is public as CVE-2026-107739, scored 7.2 of 10 by WPScan, and credited to RIA Labs.

Patched

Read this part first

If your site runs a database search-and-replace plugin, read the advisory. It names the plugin and the affected versions: before 3.2.4. The fix is in 3.2.4, dated 12 November 2026 in its changelog. Update to it or later.

This matters on a WordPress multisite network, where many sites share one install. On a single site, an administrator already holds most of the keys. Then open your list of Super Admins. Our own suggestion, not the advisory’s: look for any account you do not recognise.

WPScan is holding its proof of concept back until 16 December 2026, to give sites time to update. There is no exploit code below, and none of the names that would make one easy to write.

The motive

A detective starts with who gains. On a network, a Super Admin controls every site on it, and the network’s own tables. A subsite administrator is a tenant. Think of a university department, a client of an agency, or a customer of a host. They run their own site and nothing else.

The plugin is a database tool. It rewrites text across tables, so it needs a limit on which tables an administrator may touch. That limit is the wall between tenants. If it fails, the tenant reads and edits other sites and the network, and can make themselves Super Admin.

How bad it is

7.2 of 10, WPScan’s score: CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:H/I:H/A:H. Over the network, no click, and total loss on all three counts. Our own report said 8.7. We scored the scope as changed, because the subsite administrator reaches the whole network. WPScan scored it as unchanged. WPScan’s number is the one of record.

The precondition is real and it is the whole story: a wordpress multisite network, and an attacker who already holds the administrator role on one subsite.

The tip-off that wasn’t

Nobody called us. We did not start with a complaint. We started with a fix. Our patch-diff pass reads each security release and asks one question: did this close the hole, or one door into it?

The release in question, 3.2.2 in May 2024, was the first from the plugin’s new owner, and its changelog says it fixed a potential time-based SQL injection. WPScan lists two administrator-level SQL injection records, published a week later and marked fixed in 3.2.2. Our report treated what we found as a flaw in the fix that came with that release.

The first clue: a noisy diff

The diff of that security release ran to almost 200 KB across 21 files. It looked like a rewrite. It was not. Strip the differences that are only line endings and four real changes are left.

Noise hides fixes. A reader who skims a diff that size sees effort, and effort looks like a cure.

The second clue: the scene had not changed

The plugin has one file that does the dangerous work. Six places in it take a table name and hand it to the database. After the fix, that file was identical to the one before it, byte for byte.

So the fix was not in the room where the harm happens. It was a guard in a different file, in front. A detector that looks only at the file the fix touched never sees the six places. That is how this kind of fault stays quiet. The patch is easy to find, and it is not where the risk is.

The suspect: the guest list

The guard was a guest list. Only tables on the list may be used. That is the right idea, and an allow-list is better than a block-list. We read how it was applied.

It did not check every name it was handed. It checked the names it expected to see. A guard that looks at the list it expects is not a guard on the list it receives, and the six places behind it trusted it completely.

We are not going to show which names slip through. The vendor’s own fix says what was missing, in public: “Validate every submitted table against the available tables.” That is 3.2.4, and it is the line this post rests on.

The alibi we checked

Before we filed, we asked if anyone had got there first. The plugin had three public records on WPScan: two SQL injections from May 2024, scored 3.5 and 4.1, and an object injection from July 2024. All were fixed by 3.2.3. Nothing described a bypass of the allow-list.

We also asked if it even mattered. On a single site, an administrator can already install code, so a table bypass gains them almost nothing. The finding only counts across the wall between sites. So we built a network of two sites to test it. A claim about a network needs a network.

How we know

A description is not a finding. Our agent flagged the fix. We did not take its word. In August 2026 we ran 3.2.3 in that network of two sites. The attacker was an administrator of one subsite. Before, a request that only a Super Admin may make was refused with a 403. After, the same request answered 200, and the account was a Super Admin.

We ran it three times and got the same result each time. These are the results from our lab notes. We did not run them again for this post.

We reported it to WPScan on 15 August 2026. The fixed release came on 12 November. WPScan published four days later.

Three questions for your own fixes

None of this is specific to one plugin. Any fix that adds a list of allowed things is worth asking these about.

  1. Does the check cover everything it receives, or only what you expected? Send it more than the shape you built it for. Test the entry you did not plan for.
  2. Did the fix change the place where the harm happens? If the dangerous code is untouched, the fix is one more guard in front of it. Ask what else can reach it.
  3. Where is the wall between your tenants, and who tested it from the inside? A tool that works across tables needs a test from the weakest account allowed to run it.

None of it needs exotic tooling. It needs someone to read the fix as an attacker would, with the first fix open next to it.

What it proves, and what it doesn’t

It proves that on 3.2.3, in a lab network of two sites, an administrator of one subsite became a Super Admin.

It does not prove we found this alone or first. The 3.2.4 changelog names no one. The WPScan record lists RIA Labs as researcher. On the same day WPScan published a second record against the same release, for a different route. A different researcher is credited for it, and this post is not about it. We ran 3.2.3 only, though the affected range is wider. We have not retested 3.2.4. The listing shows 100,000+ active installs, and we cannot say how many are multisite, or how many have updated. When we looked on 16 November, the CVE registry did not yet hold the record, so the identifier above is not a link.

If you ship a plugin, or run a network full of them, we’ll read it. Nominate a target and the first scan costs nothing.

Published 16 November 2026. CVE-2026-107739 · CVSS 3.1 base score 7.2 · CWE-269 (Improper Privilege Management) · credited to RIA Labs, scored and published by WPScan. Status as of 16 November 2026. Corrections, or a plugin you maintain and want looked at: info@ria-labs.com.

See what a scan turns up

Nominate a target. The first scan costs nothing, and we’ll walk you through it.