Summary
XSS sanitization is incomplete, some attributes are missing such as oncontentvisibilityautostatechange=. This allows for the email preview to render HTML that executes arbitrary JavaScript,
Details
Sanitization is implemented here:
https://github.com/novuhq/novu/blob/next/libs/application-generic/src/services/sanitize/sanitizer.service.ts
With allowedAttributes: false, all attributes are allowed through sanitize-html. Even dangerous ones like oncontentvisibilityautostatechange=. The DANGEROUS_ATTRIBUTES array tries to handle this by denying more attributes after the fact, but this list is incomplete. I copied all well-known payloads from:
https://portswigger.net/web-security/cross-site-scripting/cheat-sheet
And found that the oncontentvisibilityautostatechange= attribute isn't detected.
PS. there seems to also be another even more lax sanitizer here, but I wasn't able to figure out where it is used:
https://github.com/novuhq/novu/blob/next/packages/framework/src/utils/sanitize.utils.ts
PoC
- Create a new workflow and add an Email step
- In the body, write the following HTML code:
<a oncontentvisibilityautostatechange="alert(window.origin)" style="display:block;content-visibility:auto">
- Wait a second and notice the XSS popup showing the current origin:
https://dashboard.novu.co/env/dev_env_gVtdgDEhgf1CetwX/workflows/onboarding-demo-workflow_wf_gVtdh2uV0h7j3ffK/steps/email-step_st_gVtqdgIrOkYVvP9F/editor
Impact
This may look like a Self-XSS similar to GHSA-w8vm-jx29-52fr, but it can be more impactful. First of all, if multiple users can access this dashboard, the link above can directly bring the to the email step editor to trigger the XSS.
An attacker can also use the Google/GitHub OAuth flows without completing the code callback step, and send that URL to the victim to intentionally log the vicitm into the attacker's account. If the attacker has prepared an XSS payload there, they will now be allowed to view it, so it triggers.
References
JorianWoltjer Reporter
Summary
XSS sanitization is incomplete, some attributes are missing such as
oncontentvisibilityautostatechange=. This allows for the email preview to render HTML that executes arbitrary JavaScript,Details
Sanitization is implemented here:
https://github.com/novuhq/novu/blob/next/libs/application-generic/src/services/sanitize/sanitizer.service.ts
With
allowedAttributes: false, all attributes are allowed throughsanitize-html. Even dangerous ones likeoncontentvisibilityautostatechange=. TheDANGEROUS_ATTRIBUTESarray tries to handle this by denying more attributes after the fact, but this list is incomplete. I copied all well-known payloads from:https://portswigger.net/web-security/cross-site-scripting/cheat-sheet
And found that the
oncontentvisibilityautostatechange=attribute isn't detected.PS. there seems to also be another even more lax sanitizer here, but I wasn't able to figure out where it is used:
https://github.com/novuhq/novu/blob/next/packages/framework/src/utils/sanitize.utils.ts
PoC
https://dashboard.novu.co/env/dev_env_gVtdgDEhgf1CetwX/workflows/onboarding-demo-workflow_wf_gVtdh2uV0h7j3ffK/steps/email-step_st_gVtqdgIrOkYVvP9F/editor
Impact
This may look like a Self-XSS similar to GHSA-w8vm-jx29-52fr, but it can be more impactful. First of all, if multiple users can access this dashboard, the link above can directly bring the to the email step editor to trigger the XSS.
An attacker can also use the Google/GitHub OAuth flows without completing the code callback step, and send that URL to the victim to intentionally log the vicitm into the attacker's account. If the attacker has prepared an XSS payload there, they will now be allowed to view it, so it triggers.
References