Meta Muse’s URL Design Exposes Sensitive User Information
Meta’s new AI assistant, Muse, has been reported to expose sensitive information when users share conversations via web links. Researchers from the security firm Claroc revealed that a single query can uncover thousands of conversation links created by Muse users.
Muse allows users to share various types of content, including conversations and images, by generating unique URLs. While these URLs are typically designed to be complex and hard to guess, the way Muse constructs them has made them easy to discover.
Easy Discovery of Sensitive Links
According to Claroc, they were able to find 6,800 links to shared conversations simply by performing a passive DNS search. This method involves tracking historical domain name resolutions, which can reveal URLs that were not intended for public access. Among the sensitive information uncovered were:
- A candidate’s notes for a confidential executive-search call, including internal contact details and meeting times.
- An internal revenue review from a national marketing company, marked for the CEO.
- A small business owner’s loan application with the requested amount.
- A to-do list containing a bank fraud alert with partial account details.
- A real estate investor’s performance report with specific property addresses and financial figures.
- A workplace’s full-year vacation calendar listing employee absences.
The researchers noted that gathering this information was easier than anticipated, stating, “If you had asked us beforehand, we would have guessed that gathering these would take real effort. It took one query.”
Design Flaw in URL Structure
The URLs for Muse’s shared content include the conversation title along with a long string of text, making them searchable. For example, links like home-sale-budget-9f21c8a3.cf.metaaiusercontent.com and loan-application-draft-a4c9e2.cf.metaaiusercontent.com are easily identifiable due to their structure. Claroc referred to these as “GhostShares,” indicating that they can be accessed by individuals who were not directly given the link.
Despite Meta CEO Mark Zuckerberg’s claims that Muse was designed with privacy and security in mind, the ease of discovering these links raises significant concerns. Muse has quickly gained popularity, becoming one of the top apps in Apple’s App Store, and it requires extensive access to users’ accounts to function effectively.
Three weeks after Muse’s launch, a zero-day vulnerability was discovered by macOS security expert Patrick Wardle, which allowed other apps to gain control of users’ Muse accounts. Meta downplayed the severity of this vulnerability, suggesting exploitation was only possible if a device was already compromised.
Claroc’s co-founder and CTO, Asaf Aprozper, stated that his company reached out to Meta regarding these findings but did not receive a response. He emphasized that users should be cautious when sharing links, advising them to limit personal information in conversations and delete links promptly.
Related link: blog.claroc.io
Source: Ars Technica