IT之家 10 月 24 日消息,Meta 旗下个人 AI 助手 Muse 的对话分享能力被曝存在隐私风险。
网络安全研究团队 Claroc 于 10 月 18 日发布报告称,研究人员借助被动 DNS(Passive DNS)查询发现,超过 6800 个 Muse 分享页面的域名可以被检索到,部分页面包含个人财务资料、企业内部文件及机密会议准备信息。Meta 随后暂停了该能力,并表示正在调查难题及评估改进思路。
Claroc 研究团队发现,Muse 会依据使用者分享的对话标题生成网页地址,并将其部署在 cf.metaaiusercontent.com 域名下。由于对话标题被写入子域名,当使用者或接收者访问分享页面时,相关域名可能被被动 DNS 服务记录,进而浮现在可查询的历史 DNS 数据库中。
研究人员表示,借助一次被动 DNS 查询,他们便发现了超过 6800 个分享页面的主机名。Claroc 随后抽查了数百个页面,发现其中部分内容涉及企业营收审查文件、小企业贷款申请、银行欺诈警报、房地产投资数据,以及员工年度休假安排等敏感信息。研究人员还发现,一名候选人为与一家前沿 AI 实验室负责人开展机密电话沟通准备的材料,其中包含内部联系人信息和会议时间。
Claroc 指出,相关页面无需登录 Meta 账号即可访问,只要可以取得对应链接即可查看内容。难题并非源于攻击者破解了随机生成的标识符,而是因为完整域名被记录在被动 DNS 数据库中,研究人员可以借助查询重新发现这些分享地址。
Claroc 于 10 月 8 日借助 Meta 的漏洞赏金计划提交了相关难题。Meta 发言人对此回应称,使用者可以分享特定的 Muse 内容,并生成可借助链接访问的页面;公司已暂停这一能力,以审查研究报告并考虑如何调整链接的可发现性。
Claroc 在 10 月 20 日更新的报告中表示,研究团队重新检查了此前可以在未登录状态下打开的 40 个分享页面,发现这些页面均已无法访问,相关主机名返回错误。IT之家注意到,已经进入被动 DNS 历史记录的域名及其对话标题仍然存在,暂停分享能力并不能彻底抹除此前留下的历史记录。
研究团队同时强调,此次调查仅覆盖特定域名模式和一种发现渠道,不能据此推断 Muse 全部分享内容的暴露规模。Claroc 表示,没有证据表明这些页面已被恶意利用,也无法确定其他人是否曾在研究团队发现之前访问过相关内容。
相关阅读:
本文来源:https://www.ithome.com/1/017/517.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
IT Home reported on October 24 that the conversation-sharing feature of Meta's personal AI assistant, Muse, has been found to pose privacy risks.
On October 18, cybersecurity research team Claroc released a report stating that researchers discovered through passive DNS queries that domains from over 6,800 Muse shared pages could be indexed, with some pages containing personal financial information, internal corporate documents, and confidential meeting preparation materials. Meta subsequently suspended this capability and stated that they are investigating the issue and evaluating potential improvements.
The Claroc research team discovered that Muse generates web addresses based on the titles of conversations shared by users, deploying them under the cf.metaaiusercontent.com domain. Since the conversation title is embedded in the subdomain, the relevant domain may be recorded by passive DNS services when a user or recipient visits the shared page, thereby appearing in queryable historical DNS databases.
Researchers said that with a single passive DNS query they discovered the hostnames of more than 6,800 pages that were being shared. Claroc subsequently spot-checked hundreds of these pages, finding that some contained sensitive material such as corporate revenue audit documents, small-business loan applications, bank fraud alerts, real-estate investment data, and staff annual leave schedules. The researchers also found materials prepared for a candidate’s confidential phone call with the head of a leading AI lab, which included internal contact details and meeting times.
Claroc noted that the relevant pages are accessible without a Meta account login, as long as the corresponding links are available. The problem does not stem from attackers cracking random identifiers, but rather from the fact that the full domain names are recorded in passive DNS databases, allowing researchers to rediscover these shared addresses through queries.
Claroc submitted the issue via Meta's bug bounty program on October 8. In response, a Meta spokesperson stated that users can share specific Muse content and generate pages accessible via links; the company has suspended this capability to review the research report and consider how to adjust link discoverability.
In an updated report on October 20, Claroc stated that the research team re-examined 40 shared pages previously accessible without logging in and found they were no longer reachable, with the associated hostnames returning errors. IT Home noted that domains already in passive DNS history records and their conversation titles persist, meaning that suspending sharing capabilities does not completely erase the previously left history.
The researchers also stressed that the investigation only covered a specific domain pattern and a single discovery channel, so the full scale of exposure for all content shared by Muse cannot be inferred from it. Claroc stated that there is no evidence that these pages have been maliciously exploited, and it is also impossible to determine whether others had accessed the content before the research team discovered it.
Related Reading:
Source: https://www.ithome.com/1/017/517.htm If there are any copyright issues, please contact us at 13951431913#139.com or submit a comment on the right.

