تقرير تحليلي

Amazon says China-linked actor's AI agents found and exploited a MONAI Label zero-day

Amazon found autonomous agents built on the open source Cairn framework breaching university servers through an unpatched flaw in a medical imaging annotation tool.

INSIGHTEX Editorial Desk

مدة القراءة: 2 د

Amazon says China-linked actor's AI agents found and exploited a MONAI Label zero-dayتقرير تحليلي

What happened

A threat actor linked to China used autonomous AI agents to find and exploit a previously unknown flaw in MONAI Label, an open source tool for AI-assisted annotation of medical images, according to Amazon, SecurityWeek reported. Amazon's threat intelligence team spotted the activity in September through an exposed, unauthenticated deployment of Cairn, a framework for AI-driven penetration testing. That instance revealed both the operator's prompts and the agents' output.

The records covered a campaign of about 48 hours against university IP addresses in East Asia outside mainland China, with the agents instructed to obtain shell access. After mapping one target's API, the agents uncovered a path traversal weakness in the application's upload feature and could write arbitrary files on the server within roughly 11 minutes. In a later session they placed a Python reverse shell in a file that runs at each interpreter start, eventually gaining root.

Amazon said it was highly confident the flaw was a zero-day. It also inferred that the actor had used the same bug to compromise a server at a second university, which the operator then tunneled into through a proxy to direct the agents at other devices. MONAI Label developers have rated the bug critical for versions 0.8.5 and earlier, where authentication is off by default. A fix will ship in the next release.

Context

MONAI Label is built for AI-assisted labelling of medical images, and Cairn is an open source framework for AI-driven penetration testing. Amazon based its zero-day judgment on the lack of any fix, issue, advisory or prior database entry for the project. It tied the activity to an unidentified Chinese-nexus group with moderate confidence, citing the prompts' language, the scripts' time zone and the likely strategic value of the targeted universities.

المصدرSecurityWeek

الموجز

أسبوع التقنية في رسالة واحدة.

التقارير التي تستحق وقتك، كل أسبوع. يمكنك إلغاء الاشتراك في أي وقت.