cve.bz

CVE-2026-53995 MLflow < 3.18.0 Path Traversal via Trace ID API

8.8 highEPSS 0.6%Fix available

High (8.8) in mlflow mlflow. EPSS 0.6% estimated probability of exploitation activity in the next 30 days. Fixed in mlflow 3.17.2.

On this page
MLflow before 3.18.0 contains a path traversal vulnerability in the FileStore tracking backend that allows unauthenticated attackers to control filesystem paths by submitting a crafted trace ID containing traversal sequences to the POST /api/3.0/mlflow/traces endpoint. Attackers can exploit FileStore.start_trace() to write trace_info.yaml, request_metadata, and tag files with attacker-controlled metadata outside the intended experiment trace directory, subject to the server process's filesystem permissions. FileStore is the default tracking backend before MLflow 3.7.0 and opt-in thereafter (the default changed to a SQLite database), so deployments in the earlier part of the affected range are vulnerable without special configuration while later ones are affected only when FileStore is explicitly selected.

Severity

CVSS 4.0

8.8 high
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:H/VA:L/SC:N/SI:N/SA:N

CVSS 3.1

8.2 high
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:H/A:L
EPSS
0.6% (unchanged from 0.6% a week ago) estimated probability of exploitation activity in the next 30 days; 51st percentile. Scored 2026-11-17. A forecast for this CVE across the internet, not for your own systems.
CISA's assessment (SSVC)
Exploitation: PoC · Automatable: yes · Technical impact: partial, assessed 2026-10-19. From CISA's Vulnrichment in the CVE List: whether it is exploited, whether exploiting it can be automated, and how much control it gives.
CWE
CWE-22 Improper Limitation of a Pathname to a Restricted Directory (Path traversal)
OWASP Top 10
A01:2025 Broken Access Control
Published
Modified
Assigned by
VulnCheck (CNA)
NVD analysis
Awaiting analysis
Record from
the NVD, synced

Fix and patched versions

  • mlflow (PyPI): fixed in 3.17.2

Ask an AI about this CVE

Show the promptExplain CVE-2026-53995 for an engineer deciding whether to act. Use https://cve.bz/api/public/cves/CVE-2026-53995?via=ask-ai-copy and https://cve.bz/cve/CVE-2026-53995?via=ask-ai-copy as the source of truth: what is affected and which versions fix it, whether it is in CISA KEV, public exploits, and its EPSS (an internet-wide 30-day estimate of exploitation activity, not a forecast for my systems). If they give no fixed versions, use the CVE record https://cveawg.mitre.org/api/cve/CVE-2026-53995 or the vendor advisory, not third-party mirrors. Tell me how to check whether I am affected and how to verify a fix. Cite the sources and say where they are uncertain.

Opens a third-party assistant; its answer is its own, not cve.bz's. The prompt holds only this CVE's id and public links.

Affected

Versions

From the CNA's record in the CVE List.

ProductAffected
mlflowbefore 3.18.0

Packages (OSV)

AdvisoryEcosystemPackageAffectedFixed in
GHSA-wqmh-qccw-pqwfPyPImlflow<3.17.23.17.2

Red Hat

Red Hat rates this Important.

Red Hat's statement

Red Hat has rated this vulnerability as Important because an unauthenticated remote client can leverage path traversal sequences in trace requests to create metadata files outside the designated storage directory. The flaw carries high integrity risk due to the lack of authentication on the trace ingestion endpoint, though exploitation is constrained by the operating system permissions assigned to the MLflow service. In Red Hat OpenShift AI, exposure is limited to deployments where the MLflow server is reachable by untrusted network actors and configured to use the local FileStore backend rather than a relational database.

Red Hat's mitigation

--- CME Mitigation Guidance (2026-10-20) ---

Disclaimer: The following mitigation solutions are provided as a courtesy, pending the application of an official patch. This content was generated utilizing AI tools in conjunction with data from https://cmetaxonomy.org. Customers are advised to independently evaluate and test these temporary mitigations within their own infrastructure.

The highest-value action is to validate client-supplied trace identifiers and enforce directory boundaries within the FileStore tracking backend before writing trace metadata: upgrade to the fixed Rhoai/Odh Training Cuda128 Torch29 Py312 Rhel/Rhoai/Odh Mlflow Rhel/Rhoai/Odh Pipeline Runtime Datascience Cpu Py312 Rhel/Rhoai/Odh Pipeline Runtime Pytorch Cuda Py312 Rhel/Rhoai/Odh Pipeline Runtime Pytorch Llmcompressor Cuda Py312 Rhel/Rhoai/Odh Pipeline Runtime Pytorch Rocm Py312 Rhel/Rhoai/Odh Pipeline Runtime Tensorflow Cuda Py312 Rhel/Rhoai/Odh Th06 Cpu Torch210 Py312 Rhel/Rhoai/Odh Th06 Cuda130 Torch210 Py312 Rhel/Rhoai/Odh Th06 Rocm64 Torch291 Py312 Rhel/Rhoai/Odh Workbench Codeserver Datascience Cpu Py312 Rhel/Rhoai/Odh Workbench Jupyter Datascience Cpu Py312 Rhel/Rhoai/Odh Workbench Jupyter Pytorch Cuda Py312 Rhel/Rhoai/Odh Workbench Jupyter Pytorch Llmcompressor Cuda Py312 Rhel/Rhoai/Odh Workbench Jupyter Pytorch Rocm Py312 Rhel/Rhoai/Odh Workbench Jupyter Tensorflow Cuda Py312 Rhel/Rhoai/Odh Workbench Jupyter Trustyai Cpu Py312 Rhel/Rhoai/Odh Pipeline Runtime Tensorflow Rocm Py312 Rhel/Rhoai/Odh Workbench Jupyter Tensorflow Rocm Py312 Rhel/Rhoai/Odh Th Torch Cpu Py312 Rhel/Rhoai/Odh Th Torch Cuda Py312 Rhel/Rhoai/Odh Th Torch Rocm Py312 Rhel/Google Lightspeed Agent builds, which validate trace IDs as safe directory names and verify resolved directory boundaries before trace operations, returning an HTTP 400 error for unsafe IDs, closing crafted trace identifier with directory traversal sequences in POST /api/3.0/mlflow/traces. Until patched, this is a concern on MLflow tracking servers configured to use the FileStore tracking backend, so restrict network access to the POST /api/3.0/mlflow/traces endpoint or migrate the tracking store from FileStore to a database backend such as SQLite. As a compensating measure, normalize and canonicalize trace IDs and file paths before security checks to prevent traversal bypasses using encoding or relative references, inspect incoming POST requests to /api/3.0/mlflow/traces for directory traversal patterns such as '../' and block the payload at the perimeter using a web application firewall, and detect anomalous file write activity performed by the MLflow process outside its designated storage directory via eBPF runtime monitoring. These controls limit blast radius and contain exploitation, but they do not fix the underlying path traversal defect — apply the upstream patch to fully remediate.

Highest-impact CME: CME-1301 Path Traversal Prevention (Canonicalization) (high)

ProductPackageStateAdvisory
Red Hat OpenShift AI (RHOAI)rhoai/odh-mlflow-rhel9Not affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-datascience-cpu-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-pytorch-cuda-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-pytorch-llmcompressor-cuda-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-pytorch-rocm-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-tensorflow-cuda-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-tensorflow-rocm-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-th06-cpu-torch210-py312-rhel9Not affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-th06-cuda130-torch210-py312-rhel9Not affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-th06-rocm64-torch291-py312-rhel9Not affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-th-torch-cpu-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-th-torch-cuda-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-th-torch-rocm-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-training-cuda128-torch29-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-codeserver-datascience-cpu-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-datascience-cpu-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-pytorch-cuda-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-pytorch-llmcompressor-cuda-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-pytorch-rocm-py312-rhel9Affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-tensorflow-cuda-py312-rhel9Affected
2 more in the JSON and on Red Hat's page.

Contains Red Hat Security Data © Red Hat, Inc., licensed under CC BY 4.0. Source: https://access.redhat.com/security/data. Modified: reformatted. Red Hat's page for CVE-2026-53995.

GitHub PoC

Repositories on GitHub named after this CVE or listed for it by the PoC-in-GitHub index. For those marked Harbor Check: matches CVE, Flawharbor read the code and found it implements this CVE's exploitation technique: read, not run, and its dependencies only checked for warning signs, so a pointer and no guarantee. The rest are unverified: some hold a working proof of concept, some nothing useful, and some malware made to look like one. Read the code before you run any of it. How Harbor Check reads them, and what it found.

Attack patterns (CAPEC)

  • CAPEC-64 Using Slashes and URL Encoding Combined to Bypass Validation Logic Likelihood: High · Severity: High
  • CAPEC-76 Manipulating Web Input to File System Calls Likelihood: High · Severity: Very High
  • CAPEC-78 Using Escaped Slashes in Alternate Encoding Likelihood: High · Severity: High
  • CAPEC-79 Using Slashes in Alternate Encoding Likelihood: High · Severity: High
  • CAPEC-126 Path Traversal Likelihood: High · Severity: Very High

The attack patterns MITRE's CAPEC relates to this record's weaknesses, as MITRE names them. Credits.

References

Check your stackJSON