Sign in
Deutsch English Español Polski
Critical

Mlflow: Command Injection when serving models with enable_mlserver=True

  • Disclosed: 31.03.2026
  • Updated: 11.10.2026 12:20

A command injection vulnerability exists in Mlflow when serving a model with enable_mlserver=True. The model_uri is embedded directly into a shell command executed via bash -c without proper sanitization. If the model_uri contains shell metacharacters, such as $() or backticks, it allows for command substitution and execution of attacker-controlled commands. This vulnerability affects the latest version of mlflow/mlflow and can lead to privilege escalation if a higher-privileged service serves models from a directory writable by lower-privileged users.

Affected software

Click a technology to see all vulnerabilities affecting it.

PlatformPackageVulnerable versionsFixed in
Python (PyPI) mlflow < 3.9.0 3.9.0

References

CVSS score 9.6/10
CVSS vector
CVSS:3.0/AV:A/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
Weakness (CWE)
CWE-78
Platform
Python (PyPI)
Package
mlflow
Disclosed
31.03.2026
Updated
11.10.2026 12:20

Sources and licenses

OSV / GitHub Advisory Database - GHSA-rvhj-8chj-8v3c - CC BY 4.0

OSV / GitHub Advisory Database - PYSEC-2026-424 - PyPA Advisory Database; CC BY 4.0

Vulnerability history: mlflow

All published vulnerabilities of this package (Python (PyPI)), newest first. Shows whether this is a one-off slip or a package that needs watching.

All (82) in the archive
Total entries
82
First → latest
02.2022 09.2026
Without a fix
18
Exploited in attacks
1

Severity breakdown

  • Critical 20
  • High 48
  • Medium 13
  • Low 1

Disclosures per year

Severity:
  1. MLFLOW_ALLOW_PICKLE_DESERIALIZATION=False safety control bypassed by mlflow.statsmodels flavor — RCE via crafted model artifact
    High · 8.8 ghsa-gqvg-gmmx-x4hm >= 2.1.0, < 3.15.0 3.15.0
  2. MLflow: Unauthenticated full-read SSRF in webhook delivery: _validate_webhook_url bypassed via unvalidated HTTP redirects (and DNS rebinding)
    Critical · 9.3 Exploited in attacks CVE-2026-64849 >= 3.3.0, < 3.15.0 3.15.0
  3. MLflow AI Gateway permits SSRF through an unvalidated api_base
    High · 7.1 CVE-2026-71211 >= 3.13.0, <= 3.15.2 no fix
  4. MLflow: trace API endpoints lack proper authorization validators
    High · 8.1 CVE-2026-8147 >= 2.14.0rc0, < 3.13.0rc0 3.13.0rc0
  5. A vulnerability in mlflow/mlflow versions prior to 3.11.0 allows for the resolution of environment variables in AI Gateway secrets, which can be exploited to exfiltrate sensitive server-side environme
    High · 7.7 CVE-2026-4035 < 3.11.0 3.11.0
  6. MLflow allows unauthorized access to multipart upload endpoints when the `--serve-artifacts` mode is enabled
    Critical · 9.0 CVE-2026-2651 < 3.10.0 3.10.0
  7. MLflow: Improper Origin Validation in MLflow Assistant /ajax-api Endpoints Enables Browser-Mediated Local Command Execution
    Critical · 9.6 CVE-2026-2611 >= 3.9.0, < 3.10.0 3.10.0
  8. MLFlow Creates a Temporary File With Insecure Permissions
    High · 7.0 CVE-2026-4137 >= 1.26.0, < 3.11.0 3.11.0
  9. A vulnerability in mlflow/mlflow versions 3.9.0 and earlier allows unauthenticated access to certain FastAPI routes when the server is started with authentication enabled (`--app-name basic-auth`) and
    High · 8.6 CVE-2026-2652 < 3.10.0 3.10.0
  10. MLflow allows an unauthenticated remote attacker to read arbitrary files from the server's filesystem
    High · 7.5 CVE-2026-2614 < 3.10.0 3.10.0
  11. A Server-Side Request Forgery (SSRF) vulnerability exists in MLflow versions prior to 3.9.0. The `_create_webhook()` function in `mlflow/server/handlers.py` accepts a user-controlled `url` parameter w
    High · 7.1 CVE-2026-2393 < 3.9.0 3.9.0
  12. mlflow: FastAPI job endpoints under `/ajax-api/3.0/jobs/*` are not protected by authentication or authorization
    Critical · 9.1 CVE-2026-0545 <= 3.10.1 no fix
  13. Mlflow: Command Injection when serving models with enable_mlserver=True this entry
    Critical · 9.6 CVE-2026-0596 < 3.9.0 3.9.0
  14. MLflow Command Injection vulnerability
    Critical · 10.0 CVE-2025-15379 < 3.8.1 3.8.1
  15. MLFlow path traversal vulnerability
    Critical · 9.6 CVE-2025-15036 < 3.9.0rc0 3.9.0rc0
  16. MLFlow allows Tracing + Assessments Access
    High · 8.1 CVE-2025-15381 <= 3.8.1 no fix
  17. Arbitrary file write via tar traversal in mlflow
    High · 8.1 CVE-2025-15031 < 3.9.0rc0 3.9.0rc0
  18. MLflow has a command injection in mlflow/sagemaker/__init__.py
    High · 7.5 CVE-2025-14287 < 3.8.0rc0 3.8.0rc0
  19. MLflow Tracking Server Artifact Handler Directory Traversal Remote Code Execution Vulnerability
    High · 8.1 CVE-2026-2033 < 3.8.0rc0 3.8.0rc0
  20. MLflow Use of Default Password Authentication Bypass Vulnerability
    Critical · 9.8 CVE-2026-2635 < 3.8.0rc0 3.8.0rc0
  21. mlflow Creates of Temporary File in Directory with Insecure Permissions
    High · 7.0 CVE-2025-10279 < 3.4.0rc0 3.4.0rc0
  22. MLFlow is vulnerable to DNS rebinding attacks due to a lack of Origin header validation
    High · 8.1 CVE-2025-14279 < 3.5.0 3.5.0
  23. MLflow Weak Password Requirements Authentication Bypass Vulnerability
    High · 8.1 CVE-2025-11200 < 2.22.0rc0 2.22.0rc0
  24. MLflow Tracking Server Model Creation Directory Traversal Remote Code Execution Vulnerability
    High · 8.1 CVE-2025-11201 < 2.22.4, >= 3.0.0rc0, < 3.0.0 2.22.4, 3.0.0
  25. MLflow has a Local File Read/Path Traversal in dbfs
    High · 7.5 CVE-2024-8859 < 2.17.0rc0 2.17.0rc0
  26. Excessive directory permissions in MLflow leads to local privilege escalation when using spark_udf. This behavior can be exploited by a local attacker to gain elevated permissions by using a ToCToU at
    High · 7.0 CVE-2024-27134 < 2.16.0 2.16.0
  27. A Local File Inclusion (LFI) vulnerability was identified in mlflow/mlflow, specifically in version 2.9.2, which was fixed in version 2.11.3. This vulnerability arises from the application's failure t
    High · 7.5 CVE-2024-2928 < 2.11.3 2.11.3
  28. A vulnerability in mlflow/mlflow version 8.2.1 allows for remote code execution due to improper neutralization of special elements used in an OS command ('Command Injection') within the `mlflow.data.h
    High · 8.8 CVE-2024-0520 < 2.9.0 2.9.0
  29. MLFlow improper input validation
    High · 8.8 CVE-2024-37061 >= 1.11.0, <= 2.13.1 no fix
  30. MLFlow unsafe deserialization
    High · 8.8 CVE-2024-37057 >= 2.0.0rc0, <= 2.14.1 no fix
  31. MLFlow unsafe deserialization
    High · 8.8 CVE-2024-37058 >= 2.5.0, <= 2.14.1 no fix
  32. MLFlow unsafe deserialization
    High · 8.8 CVE-2024-37060 >= 1.27.0, <= 2.14.1 no fix
  33. MLFlow unsafe deserialization
    High · 8.8 CVE-2024-37059 >= 0.5.0, <= 3.4.0 no fix
  34. MLFlow unsafe deserialization
    High · 8.8 CVE-2024-37054 >= 0.9.0, <= 2.14.1 no fix
  35. MLFlow unsafe deserialization
    High · 8.8 CVE-2024-37052 >= 1.1.0, <= 2.14.1 no fix
  36. MLFlow unsafe deserialization
    High · 8.8 CVE-2024-37055 >= 1.24.0, <= 2.14.1 no fix
  37. MLFlow unsafe deserialization
    High · 8.8 CVE-2024-37056 >= 1.23.0, <= 2.14.1 no fix
  38. MLFlow unsafe deserialization
    High · 8.8 CVE-2024-37053 >= 1.1.0, <= 2.14.1 no fix
  39. A path traversal vulnerability exists in mlflow/mlflow version 2.11.0, identified as a bypass for the previously addressed CVE-2023-6909. The vulnerability arises from the application's handling of ar
    High · 7.5 CVE-2024-3848 < 2.12.1 2.12.1
  40. mlflow/mlflow is vulnerable to Local File Inclusion (LFI) due to improper parsing of URIs, allowing attackers to bypass checks and read arbitrary files on the system. The issue arises from the 'is_loc
    Critical · 9.3 CVE-2024-3573 < 2.10.0 2.10.0
  41. mlflow vulnerable to Path Traversal
    High · 8.1 CVE-2024-1560 <= 2.9.2 no fix
  42. mlflow vulnerable to Path Traversal
    High · 7.5 CVE-2024-1594 <= 2.9.2 no fix
  43. mlflow vulnerable to Path Traversal
    High · 7.5 CVE-2024-1593 <= 2.9.2 no fix
  44. mlflow vulnerable to Path Traversal
    High · 7.5 CVE-2024-1558 < 2.12.1 2.12.1
  45. mlflow Path Traversal vulnerability
    High · 7.5 CVE-2024-1483 < 2.12.1 2.12.1
  46. Insufficient sanitization in MLflow leads to XSS when running a recipe that uses an untrusted dataset. This issue leads to a client-side RCE when running the recipe in Jupyter Notebook. The vulnerabil
    Critical · 9.6 CVE-2024-27133 < 2.10.0 2.10.0
  47. Insufficient sanitization in MLflow leads to XSS when running an untrusted recipe.
    Critical · 9.6 CVE-2024-27132 < 2.10.0 2.10.0
  48. MLflow Path Traversal Vulnerability
    High · 8.8 CVE-2023-6976 < 2.9.2 2.9.2
  49. MLflow Local File Disclosure Vulnerability
    High · 7.5 CVE-2023-6977 < 2.9.2 2.9.2
  50. MLflow Server-Side Request Forgery (SSRF)
    Critical · 9.8 CVE-2023-6974 < 2.9.2 2.9.2
  51. MLFlow Path Traversal Vulnerability
    Critical · 9.8 CVE-2023-6975 < 2.9.2 2.9.2
  52. mlflow Command Injection vulnerability
    High · 8.8 CVE-2023-6940 < 2.9.2 2.9.2
  53. Path Traversal: '\..\filename' in GitHub repository mlflow/mlflow prior to 2.9.2.
    High · 7.5 CVE-2023-6909 < 2.9.2 2.9.2
  54. Path Traversal: '\..\filename' in GitHub repository mlflow/mlflow prior to 2.9.2.
    High · 8.1 CVE-2023-6831 < 2.9.2 2.9.2
  55. Path Traversal in GitHub repository mlflow/mlflow prior to 2.9.2.
    High · 8.8 CVE-2023-6753 < 2.9.2 2.9.2
  56. Improper Neutralization of Special Elements Used in a Template Engine in GitHub repository mlflow/mlflow prior to 2.9.2.
    High · 8.8 CVE-2023-6709 < 2.9.2 2.9.2
  57. Information exposure in MLflow
    High · 7.5 CVE-2023-43472 < 2.9.0 2.9.0
  58. MLflow authentication requirement bypass can allow a user to arbitrarily create an account
    Critical · 9.1 CVE-2023-6014 < 2.8.0 2.8.0
  59. MLflow allowed arbitrary files to be PUT onto the server
    Critical · 10.0 CVE-2023-6015 < 2.8.1 2.8.1
  60. Remote Code Execution due to Full Controled File Write in mlflow
    Critical · 10.0 CVE-2023-6018 < 2.9.2 2.9.2
Show more (8) Showing 60 of 68

Know which vulnerabilities affect your application

With the DockRay integration and software inventory enabled, DockRay checks reported component versions against known vulnerabilities and notifies you about confirmed matches. Coverage depends on the supported technology and available advisory data.

Connect your application
DockRay newsletter

Application security in your inbox

Important vulnerabilities, actionable guidance and practical advice for your applications. Choose your technologies and follow what matters to your project.

  • A daily digest of new vulnerabilities, critical ones included.
  • Notifications about important updates.
  • Practical advice on application security and monitoring.

Leave empty to receive updates for every technology.

Privacy policy

Free to subscribe. Unsubscribe with one click in any email.