⏱️ Reading time: 6 min
One click was enough. Any website could use it to trigger a universal XSS (UXSS) in uBlock Origin and inject JavaScript into any other open tab, including a banking site or a social network. The vector was a piece of the extension almost nobody reviews: the element picker. The researcher who found it, known as loops, published the full technical writeup on November 15, 2026 (source); uBlock Origin has already fixed the bug.
📑 En este artículo
- How a website could hijack uBlock Origin’s picker
- How the opt field let an attacker smuggle a trusted filter
- Who found the bug and how it got fixed
- What this means for browser extension developers
- Why finding this bug doesn’t close the security question
- uBlock Origin versus AdGuard and Brave’s native blocking
- Limitations of this finding
- References
The flaw mainly affects the MV2 version of uBlock Origin for Firefox, the most widely installed one, and to a lesser extent uBO Lite, the MV3 variant (source). It doesn’t target a specific page; it targets the extension’s internal trust mechanism, so any website could serve as the entry point.
How a website could hijack uBlock Origin’s picker
The element picker is the window that appears when you right-click and choose to block an element on the page. Firefox loads it as an internal iframe, with a URL like moz-extension://<uuid>/web_accessible_resources/epicker-ui.html, where the UUID is random per installation and is supposed to stay secret (source). The problem is that this UUID leaks on its own.
uBlock Origin ships by default with a filter from the uAssets list that intercepts fetch() calls to certain ad domains and responds with a dummy script that reads from an internal extension URL via XHR. A malicious site can override XMLHttpRequest.prototype.open before triggering that fetch, capture the URL that passes through it, and recover the full UUID (source). With the UUID in hand, the attacker builds its own iframe pointing at the real picker and talks to it via postMessage, without the user noticing anything.
How the opt field let an attacker smuggle a trusted filter
The picker normally limits what the user can write and won’t let them create scriptlet-type filters, the ones that execute JavaScript. But one of the messages it receives carries an opt field, meant to indicate the type of resource to block, and that value gets concatenated directly into the final filter with a $ symbol in front of it (source).
If the attacker puts a newline inside opt, the picker ends up writing two filter lines instead of one, and the second one can be a trusted scriptlet pointing at any domain the attacker chooses. Filters added by the picker are marked as trusted until the next rule recompile, so the script runs with the same privileges as an official uBlock Origin filter (source). From there the attack is mechanical: set the iframe’s opacity to zero and wait for any click from the user anywhere on the page to trigger the filter creation.
Who found the bug and how it got fixed
loops found the lead with help from GPT-5.6 Sol, which he asked to review whether filter injection could lead to a UXSS. The model pointed at the right direction and the researcher verified the finding by hand; the proof-of-concept code was written by Claude (source). It’s a pattern that keeps showing up more often in vulnerability research: a model suggests where to look, a person confirms the bug is real.
The fix doesn’t complicate things much for anyone. The legitimate picker now receives a secret password from the content script in its first message and only accepts instructions if that password matches. Any iframe an external site tries to set up has no way to authenticate, so the attacker’s postMessage never reaches anywhere.
What this means for browser extension developers
If you maintain a Firefox or Chrome extension with privileged pages in web_accessible_resources (settings panels, pickers, onboarding screens), this case is a pointed reminder: any message arriving via postMessage to that page needs to be treated as if it came from an attacker, regardless of whether the URL is moz-extension:// or chrome-extension://. Two checks are worth a code review: confirm the message’s origin against a fixed list, and don’t assume a random extension ID is a secret, because there’s almost always some way for it to leak, as happened here with the uAssets filter.
For anyone who just uses uBlock Origin, the action is simple: update to the latest version available from the project itself, and if you manage extensions across an organization, confirm automatic updates are enabled on machines running Firefox with the MV2 version of uBlock Origin.
Why finding this bug doesn’t close the security question
In the Hacker News discussion about the finding, user nvme0n1p1 pointed out something rarely said out loud when an extension patches a bug: a defensive security team never proves there are no vulnerabilities left, it just stays one step ahead in a cat-and-mouse game with attackers, the same way airbags don’t prove car crashes are solved (Hacker News comment). For an extension that runs with broad permissions over almost everything a user browses, like uBlock Origin, that distinction matters. This UXSS fix closes one concrete door, but it doesn’t guarantee there isn’t a similar one in some other internal message the extension handles.
uBlock Origin versus AdGuard and Brave’s native blocking
uBlock Origin isn’t the only way to block ads and trackers, and each alternative’s design changes the available attack surface.
| Extension / browser | How it filters | Exposes a privileged UI like the picker |
|---|---|---|
| uBlock Origin (MV2, Firefox) | Filter lists with trust rules that can inject JS | Yes, the element picker via an internal iframe |
| uBlock Origin Lite (MV3) | Browser-native declarative rules, no arbitrary trusted scripts | Affected by a variant of the same bug, though less severe (source) |
| AdGuard (extension) | Its own filter lists plus content rules | Has its own picker and settings pages, with its own surface to audit |
| Brave (native browser blocking) | Blocking engine built into the browser, not an extension with its own UUID | Doesn’t depend on a web_accessible_resources page exposed by an extension |
Limitations of this finding
The attack depends on the uAssets filter that intercepts fetch() being active, which is the case by default in a standard uBlock Origin install, but it isn’t universal if someone runs a different filter list. It’s also Firefox-specific: the way the browser assigns a random per-installation UUID to each extension isn’t the same in Chrome, so the UUID-leaking mechanism doesn’t carry over unchanged.
The researcher doesn’t report any evidence of active exploitation before the fix. He found it, tested it in his own environment, and reported it responsibly together with the uBlock Origin team (source).
References
- iter.ca: Universal XSS in uBlock Origin from website-driven element picker: full technical detail of the bug, the exploitation and the fix.
- Hacker News: discussion thread about the finding.
- Comment by nvme0n1p1 on Hacker News: argument on why a patch doesn’t prove the absence of other vulnerabilities.
Featured image: Foto de Zulfugar Karimov en Unsplash
How each article is made (sources, testing and editorial criteria): How we work.
Did it work for you? Got a different error? Say so below: questions get answered and help the next reader.
Leave a comment
0 Comments