Summary

This is a reflected XSS attack which could cause arbitrary code to execute.
An internet user may create some content on the web with a specially crafted string of text.
If an NVDA user interacts with this string, they may have arbitrary code executed on their device

Patch commit(s)

Note: this fix, included in 2023.3.3, is only tested on Windows 8.1, Windows 10 and Windows 11. 2023.3 is the last NVDA version to work with Windows 7, and is considered insecure.

4e3544f

Limitations

A user must interact with crafted content.

Technical details

NVDA inserts a special string to determine how to split content for presenting an HTML browsable message.
Using this string, the HTML browsable message in NVDA can have arbitrary HTML injected into it.
When the browsable message dialog opens, the arbitrary code is executed on their device.

Proof of concept

Example XSS attack string: <a href="https://example.org/">Test__NVDA:split-here__&lt;img src="/" onerror="new ActiveXObject('WScript.shell').run('c:/windows/system32/calc.exe');"&gt;&lt;/img&gt;</a>

1. Example attack from markdown content: Test__NVDA:split-here__<img src="/" onerror="new ActiveXObject('WScript.shell').run('c:/windows/system32/calc.exe');"></img>
2. Press NVDA+k twice on the link.
3. Calculator should open.

Indicators of compromise

For developers, crafted content can be determined by the inclusion of __NVDA:split-here__.
This content may be logged on the device's NVDA log depending on NVDA settings.
However, if this is missed or hidden somehow there would be no known indicators of compromise.

Workarounds

None

Timeline

• Reported: 2024/1/18
• Fixed in 2023.3.3: 2024/1/29
  • Note: this fix is only tested on Windows 8.1, Windows 10 and Windows 11. 2023.3 is the last NVDA version to work with Windows 7, and is considered insecure.
• Advisory released: 2024/02/05

For more information

If you have any questions or comments about this advisory:

• Email us at [email protected]