CVE-2026-127202: MLflow FileStore log_outputs API Path Traversal & meta.yaml Poisoning
HERMES THREAT SCORE & ENTERPRISE MLOPS RISK
Target:MLflow Tracking Server - FileStore Backend (FileStore.log_outputs) CVSS v3.1 rates path traversal in log_outputs at 5.3 to 8.6 assuming isolated experiment tracking. Hermes Threat Score elevates CVE-2026-127202 to 86 (HIGH). In automated MLOps pipelines and enterprise AI architectures, MLflow serves as the central model artifact registry. Arbitrary directory creation and meta.yaml writes permit poisoning model lineage metadata, hijacking model loading paths, and achieving remote code execution when agentic workflows deserialize poisoned model weights or execution runtimes.
HASS AGENTIC SEVERITY & AUTONOMOUS PIPELINE CORRUPTION
Target:MLflow Model Registry & Autonomous Agent Artifact Pipeline Autonomous AI agents and LLM tool-calling orchestrators rely on MLflow to fetch fine-tuned model weights, prompts, and tool execution manifests. Path traversal breaking out of the run directory allows untrusted actors to inject rogue meta.yaml descriptors directly into the model loading tree, turning autonomous inference pipelines into arbitrary execution channels.
CVE-2026-127202: MLflow FileStore log_outputs API Path Traversal & meta.yaml PoisoningVULNERABILITY
Software platform affected by security vulnerabilities and agentic attack patterns.
๐ Why is this related? (Evidence & Provenance)
“Confirmed security vulnerability in MLflow Machine Learning Platform documented in Hermes dossier.”
- [vulnerability_report]
- [government_confirmation]CISA verified active exploitation in the wild and mandated federal remediation deadline in KEV entry. — Source: Cybersecurity & Infrastructure Security Agency (CISA): CISA Adds CVE-2026-59822 to Known Exploited Vulnerabilities Catalog (Reliability: VERY_HIGH)
1. Technical Metadata & Affected Surface
Section titled โ1. Technical Metadata & Affected Surfaceโ| Parameter | Technical Specification | Operational Impact |
|---|---|---|
| CVE Identifier | CVE-2026-127202 | Universal vulnerability identifier |
| Vendor Advisory | VulnCheck Advisory: MLflow Path Traversal in log_outputs | Weaponized exploit research disclosure |
| Patch Commit | ab5fc795eedc22348253cf687818cf2a16c15282 | Official upstream patch in Git repository |
| CVSS v3.1 Score | 8.6 (High) | CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:H/A:H |
| FIRST EPSS Score | 15.24% (96.52th percentile) | Elevated automated exploitation risk |
| CWE Classification | CWE-22 (Path Traversal) | Directory breakout in model output logging |
| Vulnerable Component | FileStore.log_outputs via /api/3.0/mlflow/runs/outputs | Port TCP 5000 (MLflow Tracking Server) |
| Affected Versions | MLflow versions prior to 3.18.0 | Systems operating with FileStore backend |
| Remediation Release | Version 3.18.0 | Official patch release |
2. Root Cause Analysis & Directory Traversal Mechanism
Section titled โ2. Root Cause Analysis & Directory Traversal MechanismโWhen MLflow tracks model runs and evaluation outputs, client SDKs (or automated agents) report outputs using the REST API:
POST /api/3.0/mlflow/runs/outputs
In the FileStore backend implementation in mlflow/store/tracking/file_store.py, the log_outputs function constructs the destination directory on disk by joining the base run path directly with the user-supplied model_id:
# Vulnerable code pattern prior to MLflow 3.18.0def log_outputs(self, run_id, outputs): run_dir = self._get_run_dir(run_id) # e.g. /data/mlruns/0/run123 for output in outputs: model_id = output.model_id # VULNERABILITY: os.path.join without path containment check or sanitization! # If model_id contains "../../../../tmp/malicious_dir", output_dir traverses out! output_dir = os.path.join(run_dir, "outputs", model_id)
# Creates the arbitrary destination directory on the server filesystem os.makedirs(output_dir, exist_ok=True)
# Writes an attacker-influenced meta.yaml file into the target directory meta_file = os.path.join(output_dir, "meta.yaml") with open(meta_file, "w") as f: yaml.dump(output.to_dictionary(), f)Because os.path.join replaces previous components when an absolute path or relative traversal sequences (../) reach the filesystem root, an attacker who supplies model_id: "../../../../../../var/www/html/exploit" or model_id: "../../../../../../etc/cron.d/job" escapes the mlruns directory entirely.
The Upstream Fix (Commit ab5fc795eedc)
Section titled โThe Upstream Fix (Commit ab5fc795eedc)โThe maintainers resolved the issue in version 3.18.0 by introducing strict path resolution verification using os.path.abspath and verifying that the target path remains strictly contained within the intended run outputs directory:
# Patched logic in MLflow 3.18.0output_dir = os.path.abspath(os.path.join(run_dir, "outputs", model_id))if not output_dir.startswith(os.path.abspath(os.path.join(run_dir, "outputs")) + os.sep): raise MlflowException( f"Invalid model_id '{model_id}': Directory traversal sequence detected.", error_code=INVALID_PARAMETER_VALUE )3. Attack Vectors & Agentic Pipeline Poisoning
Section titled โ3. Attack Vectors & Agentic Pipeline PoisoningโsequenceDiagram autonumber actor Attacker as Threat Actor / Rogue Agent participant API as MLflow Tracking Server (:5000) participant FS as FileStore Backend (file_store.py) participant Disk as Host Filesystem participant Agent as Autonomous Inference / Serving Agent
Attacker->>API: POST /api/3.0/mlflow/runs/outputs Note over Attacker,API: JSON Payload: run_id=run1, model_id=../../../../models/production API->>FS: Calls log_outputs(run_id="run1", outputs=[...]) Note over FS: os.path.join resolves outside mlruns root (CWE-22) FS->>Disk: os.makedirs() creates /models/production FS->>Disk: Writes crafted meta.yaml with malicious artifact location Note over Disk,Agent: Downstream agent polls MLflow registry for production model Agent->>Disk: Reads tainted meta.yaml and loads poisoned model weights Agent-->>Attacker: Deserialization / Hook execution delivers RCE to attackerAttack Scenarios in Enterprise AI Environments
Section titled โAttack Scenarios in Enterprise AI Environmentsโ- Model Lineage & Metadata Poisoning: Attackers overwrite metadata in shared model repositories, tricking autonomous agents into loading unvetted models or models hosted on adversary-controlled S3/HTTP endpoints.
- Arbitrary Directory Creation Denial of Service: Flooding the server with thousands of deep path traversal strings exhausts file system inodes or causes disk clutter across administrative directories.
- Chained Code Execution via Deserialization: In setups where MLflow serves models directly via
mlflow models serveor Ray Serve, pointing model artifact paths to poisoned pickle or PyTorch files results in arbitrary Python code execution upon model loading.
4. Forensic Triage & Detection Engineering
Section titled โ4. Forensic Triage & Detection EngineeringโSOCs and MLOps security teams should monitor HTTP logs for path traversal characters targeting the /outputs endpoint and audit filesystem writes performed by the mlflow server process.
title: MLflow log_outputs Path Traversal Attemptid: cve-2026-127202-mlflow-path-traversalstatus: experimentaldescription: Detects path traversal sequences in requests to the MLflow runs outputs API (CVE-2026-127202).author: Hermes Codex Cyber Threat Intelligencedate: 2026-11-16references: - https://vulncheck.com/advisories/mlflow-path-traversal-via-model-id-in-log-outputs-api - https://github.com/mlflow/mlflow/commit/ab5fc795eedc22348253cf687818cf2a16c15282tags: - attack.initial_access - attack.t1190 - attack.defense_evasion - attack.t1036logsource: category: webserverdetection: selection_endpoint: cs-method: 'POST' cs-uri-stem|endswith: '/api/3.0/mlflow/runs/outputs' selection_traversal: cs-uri-query|contains: - '..' - '%2e%2e' - '%2f' - '%5c' condition: selection_endpoint and selection_traversalfalsepositives: - None expected for standard model identifierslevel: high// Azure Sentinel / Application Gateway WAF Logs// Detects MLflow runs outputs path traversal exploitation attemptsW3CIISLog| where csMethod == "POST"| where csUriStem endswith "/api/3.0/mlflow/runs/outputs"| where csUriQuery has ".." or csUriQuery has "%2e%2e" or csUriQuery has "%2f" or csUriQuery has "%5c"| project TimeGenerated, cIP, sIP, csMethod, csUriStem, csUriQuery, scStatus, csUserAgent| order by TimeGenerated desc# 1. Audit directories created outside the intended mlruns tracking rootfind / -name "meta.yaml" -not -path "*/mlruns/*" -mtime -7 2>/dev/null
# 2. Check for path traversal characters in MLflow access logsgrep -E "POST.*mlflow/runs/outputs.*(\.\.|%2e%2e)" /var/log/mlflow/access.log
# 3. Monitor file creation events from mlflow service user via auditctlauditctl -w /data/mlruns -p wa -k mlflow_file_write5. Remediation Strategy & Mitigation Plan
Section titled โ5. Remediation Strategy & Mitigation Planโ- Immediate Upstream Upgrade (P0): Upgrade MLflow to version 3.18.0 or later. In Python environments:
Verify the installed version:
Terminal window pip install --upgrade "mlflow>=3.18.0"Terminal window mlflow --version - Transition from FileStore to Managed SQL Backend (P0): FileStore is intended primarily for local prototyping. In enterprise deployments, configure MLflow to use a robust relational database backend (
SQLAlchemyStorebacked by PostgreSQL, MySQL, or Amazon RDS) and an object store (S3, GCS, Azure Blob) for artifacts, eliminating filesystem path manipulation vulnerabilities. - Enforce Authentication & Reverse Proxy Guardrails (P1): Never expose MLflow tracking servers directly to the public internet without an authenticating reverse proxy (Nginx, Traefik, OAuth2 Proxy) and role-based access control.
- Isolate MLflow Service Permissions (P1): Ensure the OS user running the MLflow tracking service has write access exclusively to the designated
mlrunsdirectory and cannot write to/etc, web roots, or cron directories.