Логотип exploitDog
Консоль
Логотип exploitDog

exploitDog

redhat логотип

CVE-2026-127202

Опубликовано: 12 нояб. 2026
Источник: redhat
CVSS3: 5.3

Описание

MLflow before 3.18.0 using the FileStore backend contains a path traversal vulnerability that allows unauthenticated remote attackers to write files outside the intended run directory by supplying a malicious model_id parameter to the log outputs API. Attackers can exploit the unvalidated model_id containing ../ sequences via the POST /api/3.0/mlflow/runs/outputs endpoint to create directories and write a meta.yaml file at arbitrary filesystem locations, limited only by the server account's permissions. FileStore was the default backend before 3.7.0 and is opt-in afterward through MLFLOW_ALLOW_FILE_STORE.

A flaw was found in MLflow when using the FileStore backend. An unauthenticated remote attacker can exploit a path traversal vulnerability by submitting directory traversal sequences in the model identifier parameter to the log outputs application programming interface (API). Successful exploitation allows the attacker to create directories and write files outside the intended storage location on the server filesystem, restricted only by the permissions of the application process.

Отчет

This vulnerability is rated as Moderate severity because, although unauthenticated remote attackers can invoke the path traversal flaw, the operational impact is confined to creating directories and writing structured metadata files within the local privileges of the running service account. Red Hat deployments typically leverage relational database or object storage backends rather than the legacy FileStore, which requires explicit opt-in enablement in current releases. Because attackers cannot inject arbitrary code or extract file contents through this mechanism, overall system confidentiality and availability are preserved.

Меры по смягчению последствий

Disable the legacy FileStore backend and restrict access to the MLflow tracking service.

  1. Avoid the FileStore backend: Ensure the MLFLOW_ALLOW_FILE_STORE environment variable is unset in the MLflow server execution environment:
unset MLFLOW_ALLOW_FILE_STORE

Configure the tracking server with an alternative supported backend, such as a relational database store using the --backend-store-uri option. 2. Restrict network reachability: Limit network access to the MLflow service port (default TCP 5000) to authorized hosts using firewalld:

firewall-cmd --permanent --remove-port=5000/tcp firewall-cmd --permanent --zone=trusted --add-source=<TRUSTED_IP_RANGE> firewall-cmd --reload

Caveats: Migrating to a database backend requires moving existing experiment data and tracking metadata. Modifying service configurations requires restarting the MLflow daemon, and reloading firewall rules will temporarily disrupt existing network sessions.

Затронутые пакеты

ПлатформаПакетСостояниеРекомендацияРелиз
Red Hat OpenShift AI (RHOAI)rhoai/odh-mlflow-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-datascience-cpu-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-pytorch-cuda-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-pytorch-llmcompressor-cuda-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-pytorch-rocm-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-tensorflow-cuda-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-tensorflow-rocm-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-th06-cpu-torch210-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-th06-cuda130-torch210-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-th06-rocm64-torch291-py312-rhel9Fix deferred

Показывать по

Дополнительная информация

Статус:

Moderate
Дефект:
CWE-22
https://bugzilla.redhat.com/show_bug.cgi?id=2572655mlflow: mlflow: Arbitrary file write via path traversal in model_id parameter

5.3 Medium

CVSS3

Связанные уязвимости

CVSS3: 5.3
nvd
10 дней назад

MLflow before 3.18.0 using the FileStore backend contains a path traversal vulnerability that allows unauthenticated remote attackers to write files outside the intended run directory by supplying a malicious model_id parameter to the log outputs API. Attackers can exploit the unvalidated model_id containing ../ sequences via the POST /api/3.0/mlflow/runs/outputs endpoint to create directories and write a meta.yaml file at arbitrary filesystem locations, limited only by the server account's permissions. FileStore was the default backend before 3.7.0 and is opt-in afterward through MLFLOW_ALLOW_FILE_STORE.

CVSS3: 5.3
github
10 дней назад

MLflow before 3.18.0 using the FileStore backend contains a path traversal vulnerability that allows unauthenticated remote attackers to write files outside the intended run directory by supplying a malicious model_id parameter to the log outputs API. Attackers can exploit the unvalidated model_id containing ../ sequences via the POST /api/3.0/mlflow/runs/outputs endpoint to create directories and write a meta.yaml file at arbitrary filesystem locations, limited only by the server account's permissions. FileStore was the default backend before 3.7.0 and is opt-in afterward through MLFLOW_ALLOW_FILE_STORE.

5.3 Medium

CVSS3