Meta’s Muse AI 0-Day Lets Hackers Hijack Dictation Traffic and Inject Malicious Prompts
A local zero-day in Meta's Muse AI lets same-user malware redirect dictation and inject prompts.
Researcher Patrick Wardle published a proof-of-concept for a local zero-day in Meta’s Muse AI app, documented in the not-a-mused repository. A process running as the logged-in user, without administrator rights, can change the undocumented endo_voyager_dictation_endpoint value and send dictation traffic to an attacker-controlled service. Intercepted traffic may include voice audio and Muse authentication material, and crafted replies can be processed as prompt injection. It is not remote code execution; the attacker must already be able to run code in the user account.
- Unprivileged local code can change Muse's dictation endpoint setting.
- Redirected traffic can expose dictated audio and authentication tokens.
- Crafted service responses may be processed as prompt injection.
- Not remote code execution; prior execution in the user account is required.
- Defenders should watch config changes and unexpected Muse destinations.
From the original article
2 min read at gbhackers.com
A recent proof-of-concept (PoC) developed by security researcher Patrick Wardle reveals a local zero-day vulnerability in the Muse application. This vulnerability allows malware running under the logged-in user’s account to redirect dictation traffic to a server controlled by an attacker. Meta’s Muse AI 0-Day The issue is documented in Wardle’s “not-a-mused” research repository.
By Divya · gbhackers.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.