Local code execution via writable model cache
Published May 18, 2026 · Updated May 18, 2026
Insecure temporary-directory permissions in MLflow before 3.11.0 allow local users to execute code through tampered model artifacts. The get_or_create_nfs_tmp_dir and _create_model_downloading_tmp_dir functions chmod temporary directories to 0777 and 0770, allowing model files to be replaced before cloudpickle.load deserializes them. Exploitation requires local access to a shared filesystem or group-writable cache, and successful tampering runs code with the MLflow process privileges.
Summary
What happened
Insecure temporary-directory permissions in MLflow before 3.11.0 allow local users to execute code through tampered model artifacts. The get_or_create_nfs_tmp_dir and _create_model_downloading_tmp_dir functions chmod temporary directories to 0777 and 0770, allowing model files to be replaced before cloudpickle.load deserializes them. Exploitation requires local access to a shared filesystem or group-writable cache, and successful tampering runs code with the MLflow process privileges.
The record
- CVE
- CVE-2026-4137
- Published
- May 18, 2026
- Updated
- May 18, 2026
- Vendor
- MLflow Project
- Product
- MLflow
- Classifications
- CWE-378, T1574.010
- Attack vector
- local
- Privileges
- Unavailable
Timeline
How it unfolded
- May 18, 2026CVE publishedPublication date reported by the CVE source.
- May 18, 2026Record updatedLatest update available in the CVE record.
Exploitability
Present is not the same as exploitable
Compare your product and version with the public record. A matching version still requires validation against your environment.
Is a vulnerable build present?
Compare these published version ranges with your installed build and any vendor patches.
- Affected versionversion=unspecified <3.11.0
What conditions does exploitation require?
What is affected?
Attacks
What attackers are doing with it
Daily unique IPs observed by Shadowserver honeypots for known exploited vulnerabilities (KEVs). Missing observations do not establish an absence of attacks.
Weakness, pattern, technique
Public exploit references
- huntr proof of concept for writable MLflow model cacheproof of concept · unverified
Labels summarize the accepted research assessment. They do not indicate a test against your environment.
Technologies
Your stack
See the directory against your own environment.
Your stack
Check the software in your environment
Book a demo to see how Hinoki identifies affected software and validates exploitability in your environment.
Book a demo