Skip to content
audiotale.net
Sign in
NewsNews

uBlock Origin GHSA-cmqw-9r2w-6wv2: element picker XSS fixed in 1.75.4

A web page could drive uBlock Origin's element picker and add trusted filters with one click; the High-severity flaw is patched, so check you run 1.75.4 or newer.

audiotale.net team

· 6 min read

Written by AI from the sources linked below. How we use AI

On this page
  1. Key points
  2. What to do
  3. What happened
  4. Background
  5. How it works
  6. What it means
  7. What we don't know yet
  8. What to watch next
  9. Sources

A security researcher published details on 15 November 2026 of a one-click universal cross-site scripting (XSS) flaw in uBlock Origin, the popular ad blocker. The extension's maintainer published the matching advisory, GHSA-cmqw-9r2w-6wv2, rated High, the same day. The advisory lists uBlock Origin 1.75.4 as the patched version, and the latest stable release at the time of writing is 1.76.4, so anyone on a current version is covered.

Key points

  • Advisory GHSA-cmqw-9r2w-6wv2, "Universal XSS from website-driven element picker", was published on 15 November 2026 with severity High and no CVE listed.
  • A web page could embed uBlock Origin's element picker in an invisible iframe, drive it, and, if the user clicked anywhere on the page, add arbitrary trusted filters that run JavaScript on any site.
  • The advisory lists versions 1.75.2 and lower as affected and 1.75.4 as patched; the researcher reported it on 1 November and the final hardening shipped in 1.76.0 on 8 November.
  • The researcher, who uses the name syvb, says the main case is the Firefox (MV2) version; a similar but less severe issue touched uBlock Origin Lite.
  • The advisory lists no workarounds, so the action is to check your version and update.

What to do

  • Check your version. In Firefox open about:addons, select uBlock Origin and look at the version; in Chrome or Edge open chrome://extensions or edge://extensions with developer mode on. You want 1.75.4 or newer. At the time of writing the latest stable release listed on GitHub is 1.76.4 (15 November 2026).
  • Update if you are older. Browser stores normally update extensions on their own, but you can force a check from the same page (Firefox: the gear menu, "Check for Updates").
  • Lite users. The researcher says a similar, less problematic issue affected uBlock Origin Lite. The advisory covers uBlock Origin and gives no Lite version, so we cannot give you a fixed number for it; update it as well.
  • If you were on an old version. Open uBlock Origin's dashboard, then "My filters", and look for filters you do not recognise, particularly ones containing ##+js( that you never wrote. The advisory does not describe how to detect an attack, so this is our suggestion, not an official procedure.

What happened

The element picker is the feature that lets you right-click a page and choose an element to hide. It is built from a page inside the extension (epicker-ui.html) that the extension injects as an iframe, and that page can ask the extension's background script to add filters.

The advisory says a webpage could embed that same picker in its own iframe and control it. When the user then clicks anywhere on the page, the attacker can add arbitrary trusted filters. Filters made through the picker are treated as trusted until the filter set is recompiled, and trusted filters can inject JavaScript into pages. A website could therefore plant a filter for a different domain and run code in that domain, which is why the advisory calls it a universal XSS.

The research write-up by syvb, published on iter.ca on 15 November, explains the chain in four steps:

  1. Find the picker's address. In Firefox, the address contains a random per-install extension ID. The write-up shows that a default filter for a Google ad script makes the extension fetch a web-accessible resource at an address containing that ID, so a malicious site can hook its own XMLHttpRequest and read the ID. The write-up adds that there are other ways the ID could leak. The advisory adds that on MV2 Chromium a related secret could be extracted the same way.
  2. Embed the picker. The attacker's page creates an invisible iframe pointing at the picker.
  3. Drive it. The page sends messages to the picker. A field normally holding the request type, such as image, is appended to the filter after a dollar sign; because the attacker controls it, a newline lets them add a second, entirely separate filter, including a scriptlet that injects code.
  4. Get a click. The page makes the picker's "Create" button invisible and positions it so the user's click lands on it. The researcher did not build this last step in the proof of concept but says it is straightforward.

Background

The advisory gives a short fix timeline:

  • 1 November 2026: the issue is reported privately by email.
  • 2 November (version 1.75.2): a commit stops scriptlets from using potentially tainted XHR methods. The reporter then found a bypass.
  • 3 November (1.75.4): the picker launches only with a verified secret token, one of the fixes the write-up describes.
  • 8 November (1.76.0): the picker validates a filter before creating it, and filters tied to web-accessible resources are blocked "for the time being". The 1.76.0 release notes list "Universal XSS from website-driven element picker" as a security item.

The advisory names 1.75.4 as the patched version even though two of the four related commits shipped in 1.76.0; if you want every listed change, run 1.76.0 or later. Later releases, 1.76.2 and 1.76.3b0 (13 November) and 1.76.4 (15 November), carry more security-labelled commits, among them one that removes unused web-accessible-resource secrets, but we found no statement that they fix a separate vulnerability.

The researcher says an AI model, GPT-5.6 Sol, found the weak point after being asked to look for filter-injection problems, that they then verified it themselves, and that Claude wrote most of the proof-of-concept code. That is the researcher's own account; the advisory credits syvb only as reporter.

How it works

Trusted filters exist because some blocking rules need to run code in the page, for example to neutralise a tracking script. uBlock Origin lets only filters from its own lists or created by the user have that power. The element picker sits on the border: it is a user feature, so what it creates is trusted, but it also runs in an iframe that a web page can load. The original design assumed only the extension's own content script would talk to it. The fix checks that assumption with a secret the real picker receives and a web page cannot read, and independently validates the filter text.

What it means

For most users the risk window has closed: the first fix was out on 2 November, and the advisory names 1.75.4 as the patched version. People who disable updates, run old packaged builds or use browsers that lag behind the store are the exposed group. The attack needs a click on a malicious page, but because it works on any site, the consequences of a single mistaken click could include code running on banking or email pages. We have seen no report of attacks in the wild; the advisory does not mention any.

For extension developers the lesson is that a page inside the extension that is listed as web-accessible can be framed by any website, so messages into it need authentication.

What we don't know yet

  • Whether any attacker used the flaw before it was fixed. Neither the advisory nor the write-up says so.
  • The exact patched version for uBlock Origin Lite, and how its variant differs.
  • Whether other Chromium builds or the MV3 build need a separate update; the advisory only names MV2 Chromium for the secret.
  • Why the advisory dates 1.75.4 as patched when one of its own bypass fixes landed in 1.76.0.

What to watch next

  • New uBlock Origin releases after 1.76.4 and any follow-up advisory.
  • Whether other filter-based blockers publish similar fixes.
  • The maintainer's repository and the researcher's blog for any further details.

Sources

Written by

audiotale.net team

The official account of the people who run audiotale.net: announcements, conversation starters and the newsroom's drafts. Posts written with AI say so. Every other account here is a real person.