Hackers Used Cairn AI Agents to Exploit MONAI Label Zero-Day
┌─ cairn-ai-agents-exploit-monai-label.md :: hero
Amazon Threat Intelligence reported that an unidentified Chinese-nexus threat actor used autonomous AI agents from the open-source Cairn penetration testing framework to find and exploit a critical zero-day vulnerability in MONAI Label, an open-source medical image annotation tool. Detected in September during a roughly 48-hour campaign against universities in East Asia outside mainland China, the path traversal flaw affects MONAI Label versions 0.8.5 and earlier. Authentication is disabled by default, allowing network-reachable attackers to write arbitrary files and gain root shell access. Developers have created a fix for the next release.
Key Facts
| Attribute | Value |
|---|---|
| Affected Product and Versions | MONAI Label (0.8.5 and earlier) |
| Exploited in the Wild | Yes (detected in September) |
| Threat Actor | Unidentified Chinese-nexus actor (moderate confidence) |
How Did the AI Agents Discover and Exploit the Vulnerability?
The threat actor gave autonomous Cairn AI agents a target list and instructions to obtain shell access. After identifying MONAI Label on a university server and mapping its API, the agents located a path traversal bug in a file upload function, wrote arbitrary files within about 11 minutes, and later achieved root shell access.
In about 11 minutes, the agents went from mapping the application’s API to arbitrarily writing files on the server.
Amazon
In a subsequent session, the agents exploited the arbitrary file write capability to plant a Python reverse shell inside a file that executes whenever the Python interpreter starts, giving the attackers root access. Amazon Threat Intelligence assessed with high confidence that the flaw was a zero-day because no fix, issue, advisory, or prior mention existed in vulnerability databases for the open-source project.
The operator also supplied the agents with a proxy tunnelling into a second university to target other devices on its network. Because the server hosting the proxy ran the same vulnerable MONAI Label software, Amazon inferred the actor used the zero-day to compromise that host and establish the tunnel.
Autonomous Cairn agents progressed from mapping the MONAI Label API to writing arbitrary server files in about 11 minutes and later established root shell access via a Python startup file.
How Did Amazon Detect and Attribute the Campaign?
Amazon Threat Intelligence detected the campaign in September after discovering an exposed instance of Cairn, an open-source AI penetration testing framework. Because the instance lacked authentication, Amazon observed the operator’s prompts and the agents’ output across a roughly 48-hour operation targeting universities in East Asia outside mainland China.
Amazon attributed the activity to an unidentified Chinese-nexus threat actor with moderate confidence. Analysts based this assessment on three indicators observed in the exposed environment: the language used in the operator’s prompts, the time zone configured in the scripts, and the targeting of universities likely of strategic interest to China.
Amazon attributed the 48-hour university campaign with moderate confidence to an unidentified Chinese-nexus actor based on prompt language, script time zones, and target selection.
Why Is MONAI Label Vulnerable to Remote Exploitation?
MONAI Label versions 0.8.5 and earlier ship with authentication turned off by default, allowing anyone who can reach the server over the network to exploit the critical path traversal flaw. Developers have created a patch for the issue, though it will only arrive in the next software release.
The MONAI Label developers rated the vulnerability critical in an advisory published this week. Because authentication is off by default, any user or automated agent with network reachability to an unpatched server running version 0.8.5 or earlier can access the vulnerable API and trigger arbitrary file writes.
Default-disabled authentication in MONAI Label 0.8.5 and earlier allows any network-reachable attacker to exploit the critical file-upload path traversal vulnerability.
What Should You Do?
- Check whether your organisation operates MONAI Label version 0.8.5 or earlier; no fixed release is available to download yet, though developers have created a patch for the next release.
- Restrict network access to MONAI Label servers immediately and avoid exposing instances without authentication, since authentication is disabled by default.
- Inspect servers running affected versions for unauthorised files, unexpected modifications to Python interpreter startup files, or unknown proxy tunnels.
- Apply the official MONAI Label update as soon as the next release is published by the developers.
- Organisations in Malaysia that identify signs of compromise can report incidents to Cyber999 (MyCERT).
Common Questions
Is a patched version of MONAI Label available to download?
No fixed release is available to download yet. According to SecurityWeek, the MONAI Label developers published a critical advisory this week confirming the flaw affects versions 0.8.5 and earlier, and stated that a fix has been created and will be included in the next release.
What is Cairn and how was it used in this incident?
Cairn is an open-source AI penetration testing framework. In this incident, an operator gave Cairn AI agents a list of university IP addresses and instructed them to obtain shell access, enabling the agents to map MONAI Label’s API and discover the zero-day flaw.
Were organisations in Malaysia targeted in this campaign?
The SecurityWeek report does not name Malaysia or any specific victim organisations. Amazon stated that the roughly 48-hour campaign targeted IP addresses belonging to universities in East Asia outside mainland China, and noted that a proxy later tunnelled into a second university.
Sources and Methodology
This dispatch is based solely on the SecurityWeek report by Eduard Kovacs (no publication date was included in the provided source text). No external advisories, vulnerability databases, or vendor pages were reviewed or independently verified.
$ ls -t news/ | head -5
- CISA Adds Exploited VMware, Cisco, Apache, and SAP Flaws to KEV
- Broadcom Patches VMware vCenter Flaw Exploited as Zero-Day
- Kazakhstan Hackers Use AI to Spy on Central Asian Governments
- Security Researchers Identify Qilin Ransomware Administrator
- National Cyber Director Sean Cairncross Leaves White House Post
$ ls ./tags
- #Vulnerabilities
- #Nation-State
- #AI Security