---
id: "CVE-2025-15379"
page: "https://infrastack.ai/live/CVE-2025-15379"
kind: "vulnerability"
aliases: ["CVE-2025-15379", "GHSA-r23q-823p-vmf7"]
title: "MLflow Command Injection vulnerability"
severity: "critical"
cvss: 10.0
epss: 0.02401
exploited: false
status: "open"
category: "package"
products: [{"name": "mlflow", "fixed": ["3.8.1"], "affected": "< 3.8.1", "ecosystem": "pypi"}, {"name": "lfprojects mlflow", "fixed": ["3.8.1"], "affected": ">= 3.8.0, < 3.8.1", "ecosystem": "vendor"}, {"via": "cpe", "name": "mlflow", "fixed": ["3.8.1"], "affected": ">= 3.8.0, < 3.8.1", "ecosystem": "pypi"}]
ecosystems: ["pypi", "vendor"]
tags: ["cwe-77", "cwe-78", "ml-platform", "mlflow", "model-serving", "package", "python"]
not_affected_if: ["default-config", "trusted-artifacts-only"]
published_at: "2026-03-30T08:16:15+00:00"
first_seen_at: "2026-11-16T15:32:33+00:00"
updated_at: "2026-11-16T15:37:59+00:00"
sources: [{"source": "ghsa", "url": "https://github.com/advisories/GHSA-r23q-823p-vmf7", "first_seen_at": "2026-11-16T15:32:33+00:00"}, {"source": "nvd", "url": "https://nvd.nist.gov/vuln/detail/CVE-2025-15379", "first_seen_at": "2026-11-16T15:37:59+00:00"}]
exploit_sources: [{"type": "poc", "url": "https://huntr.com/bounties/dc9c1c20-7879-4050-87df-4d095fe5ca75"}, {"type": "epss", "score": 0.02401, "url": "https://api.first.org/data/v1/epss?cve=CVE-2025-15379"}]
---

# MLflow Command Injection vulnerability

Link for people: https://infrastack.ai/live/CVE-2025-15379 (this .md is for agents)

MLflow builds a shell command from dependency entries in a model's python_env.yaml without sanitizing them when deploying with env_manager=LOCAL. A malicious model artifact can make the server run arbitrary commands on any system that deploys it.

_Summary, check and fix written by a model from the sources; labelled as such._

## Affected

- pypi:mlflow affected < 3.8.1 fixed 3.8.1
- vendor:lfprojects mlflow affected >= 3.8.0, < 3.8.1 fixed 3.8.1
- pypi:mlflow affected >= 3.8.0, < 3.8.1 fixed 3.8.1

## How to check

- Check the installed version with `pip show mlflow`, looking for `3.8.0`.
- Look for model deployments that use `env_manager=LOCAL` and load artifacts from sources you do not fully control.

## Fix

- `mlflow` upgrade to `3.8.1`

## Timeline

- 2026-11-16T15:32:33+00:00 ghsa https://github.com/advisories/GHSA-r23q-823p-vmf7
- 2026-11-16T15:37:59+00:00 nvd https://nvd.nist.gov/vuln/detail/CVE-2025-15379

## Documents

- [advisory] Third-party advisory: access.redhat.com https://access.redhat.com/security/cve/CVE-2025-15379
- [advisory] Third-party advisory: bugzilla.redhat.com https://bugzilla.redhat.com/show_bug.cgi?id=2452949
- [advisory] Third-party advisory: security.access.redhat.com https://security.access.redhat.com/data/csaf/v2/vex/2025/cve-2025-15379.json
- [fix] Patch: github.com https://github.com/mlflow/mlflow/commit/361b6f620adf98385c6721e384fb5ef9a30bb05e
- [fix] Patch: github.com https://github.com/mlflow/mlflow/commit/a22ce7157f646bdce4c95106fc38ccc9ca289205
- [poc] Exploit: huntr.com https://huntr.com/bounties/dc9c1c20-7879-4050-87df-4d095fe5ca75

## From the sources

A command injection vulnerability exists in MLflow's model serving container initialization code, specifically in the `_install_model_dependencies_to_env()` function. When deploying a model with `env_manager=LOCAL`, MLflow reads dependency specifications from the model artifact's `python_env.yaml` file and directly interpolates them into a shell command without sanitization. This allows an attacker to supply a malicious model artifact and achieve arbitrary command execution on systems that deploy the model. The vulnerability affects versions 3.8.0 and is fixed in version 3.8.1.

---

_Free to read, with fair use of about 60 pages a minute. For higher limits or API access: sales@infrastack.ai_
