CVE-2025-15379
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 3.8.2.
Scoring
- Severity
- CRITICAL
- CVSS base score
- 10
- CVSS vector
- CVSS:3.0/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
- EPSS probability
- 2.39%
- CWE
- CWE-77
- Published
- 2026-03-30
- Last modified
- 2026-09-07
Affected products
- mlflow mlflow/mlflow
Weakness type
Related vulnerabilities
- CVE-2026-47670 — DbGate Vulnerable to Authenticated Remote Code Execution via loadReader functionName code injection
- CVE-2026-86152 — Tenda CP3 Kylin AutoAddWifi.cpp ThreadProc os command injection
- CVE-2026-82971 — QVidium Opera11 CGI Script net_tr.cgi command injection
- CVE-2026-72869 — Dokploy: Authenticated OS command injection in backup.restoreBackupWithLogs (databaseName) leading to host RCE
- CVE-2026-72736 — Dokploy: OS Command Injection in registry credential testing and Swarm cluster management → HOST RCE
- CVE-2026-72735 — Dokploy: Command injection in writeTraefikConfigRemote via shell interpolation of unescaped YAML in SSH remote execution
- CVE-2026-54680 — Logging operator has Fluentd configuration injection that allows remote code execution
- CVE-2026-28672 — Apache Ranger: OS Command Injection via Username in UnixUserGroupBuilder