Universal XSS in uBlock Origin from website-driven element picker
A newly disclosed universal XSS vulnerability in uBlock Origin’s element picker exposed how a malicious website could execute code inside the extension and even leak its per-installation UUID, raising both tracking and security concerns for one of the web’s most trusted ad blockers. Commenters note that the bug has been patched in recent versions but see it as evidence of the inherent risk in powerful browser extensions that inject UI into page DOMs, while still arguing that robust in-browser ad blocking remains essential given today’s ad- and malware-heavy web. The thread also touches on limits of DNS-only blocking, the need for better browser-native APIs for secure extension UI, and the growing role of AI tools in finding and exploiting such flaws.
Risk Profile of uBlock Origin & Browser Extensions
- uBO necessarily has broad access to inspect/modify pages, so bugs can have large blast radius; some see this as proof that powerful extensions are “powder kegs.”
- Others argue browsers themselves routinely ship severe RCE bugs, so focusing solely on extension XSS is inconsistent.
- General concern that extensions run complex, privileged code (including WASM) and can be hijacked, sold, or become spyware.
Element Picker UI & Browser Architecture
- Core issue: extensions often inject their UI into the page DOM (e.g., element pickers, password popups), letting hostile pages interact with extension internals.
- Several comments say the real fix is a proper browser-level “overlay” or chrome UI API for extensions, not forbidding page modification.
- Old XUL-based Firefox could place UI outside content (chrome://, overlay pages), but had its own serious risks.
UUID Leakage & Fingerprinting
- The attack involves leaking a per-install extension UUID, raising fears of unique, stable tracking IDs.
- Discussion notes Chromium has (or had) static IDs for web-accessible resources; Firefox uses random IDs, but extensions can still accidentally expose them.
- Some point out that if sites can’t enumerate all extensions/IDs, tracking utility is limited; others worry that any stable ID increases fingerprinting surface.
Mitigations, Updates & Workarounds
- Multiple reminders to manually update uBO; patched at ≥1.75.4, with some users reporting needing two updates to reach later versions.
- Suggested mitigations include blocking third‑party iframes by default and using stricter blocking modes, though it’s unclear if this would fully prevent this specific exploit.
- Android Firefox users report that removing and reinstalling uBO is currently the only practical way to force an update, at the risk of losing settings.
Ad Blocking vs DNS Blocking & NoScript
- DNS-only blocking is described as insufficient for modern, script-heavy platforms (YouTube, major social sites); others accept that trade-off because they barely use such sites.
- Some combine DNS-level blocking with uBO or similar tools; others prefer NoScript plus built-in ad-blocking in alternative browsers, despite breakage and potential hidden bugs.
Password Managers & Other Extensions
- Injected password-manager UIs can leak which accounts exist or break page behavior (e.g., 1Password’s Prism.js incident).
- Shadow DOM + iframes are discussed as a containment mechanism; some insist closed shadow roots in isolated worlds are safe from page scripts, others emphasize Shadow DOM is not a strong security boundary in general.
AI Assistance & Alignment Debate
- The blog’s “human prose, AI-generated PoC” disclosure is praised as honest and useful.
- Some are uneasy that such a “signal of trust” might itself be mimicked by future models.
- Debate over whether alignment and blue-teaming can really prevent models from autonomously faking such disclaimers; views range from “simple cases are solved” to strong skepticism, citing jailbreaks and hallucinations.