Apache Airflow: Disclosure of secrets to UI via kwargs
Description
When a DAG failed during parsing, Airflow’s error-reporting in the UI could include the full kwargs passed to the operators. If those kwargs contained sensitive values (such as secrets), they might be exposed in the UI tracebacks to authenticated users who had permission to view that DAG.
The issue has been fixed in Airflow 3.1.4 and 2.11.1, and users are strongly advised to upgrade to prevent potential disclosure of sensitive information.
AI Insight
LLM-synthesized narrative grounded in this CVE's description and references.
Airflow UI error tracebacks could expose sensitive operator kwargs (including secrets) to authenticated users viewing a failed DAG parse.
Vulnerability
Overview
CVE-2025-65995 is an information disclosure vulnerability in Apache Airflow. When a Directed Acyclic Graph (DAG) fails to parse, Airflow's error-reporting mechanism in the UI can include the full kwargs (keyword arguments) passed to the operators. If those kwargs contain sensitive values—such as secrets, passwords, API keys, or connection strings—they may be exposed in the resulting UI tracebacks [1][4].
Exploitation & Attack Surface
The vulnerability does not require an attacker to have access to logs or a command-line interface; the exposure occurs directly within the Airflow web UI. To exploit this, an authenticated user must have permission to view the DAG whose parsing failed. No additional privileges or network access beyond the standard UI are needed. The bug is rooted in how Airflow's error-handling code constructs traceback messages: it previously failed to redact operator arguments before rendering them in the UI [3].
Impact
An authenticated user who can view a DAG that encountered a parse error can potentially read any secret values passed as kwargs to the operators defined in that DAG. This could lead to credential theft, unauthorized access to external systems, or further compromise of the Airflow environment. The disclosure is limited to the UI traceback and only applies to DAGs that actually fail parsing; successfully parsed DAGs are not affected.
Mitigation & Remediation
The issue has been addressed in Airflow versions 3.1.4 and 2.11.1. The fix ensures that operator kwargs are properly redacted (masked) in error tracebacks displayed in the UI [1][3][4]. Users running earlier versions are strongly advised to upgrade to one of the patched releases. No workarounds are mentioned in the references, and the vulnerability is not currently listed in CISA's Known Exploited Vulnerabilities (KEV) catalog.
AI Insight generated on May 19, 2026. Synthesized from this CVE's description and the cited reference URLs; citations are validated against the source bundle.
Affected packages
Versions sourced from the GitHub Security Advisory.
| Package | Affected versions | Patched versions |
|---|---|---|
apache-airflowPyPI | < 2.11.1 | 2.11.1 |
apache-airflowPyPI | >= 3.0.0b1, < 3.1.5rc1 | 3.1.5rc1 |
Affected products
1- Apache Software Foundation/Apache Airflowv5Range: 3.0.0
Patches
0No patches discovered yet.
Vulnerability mechanics
AI mechanics synthesis has not run for this CVE yet.
References
6- github.com/apache/airflow/pull/58252ghsapatchWEB
- github.com/apache/airflow/pull/61883ghsapatchWEB
- github.com/advisories/GHSA-gfw7-2v73-69wgghsaADVISORY
- lists.apache.org/thread/1qzlrjo2wmlzs0rrgzgslj2pzkor0dr2ghsavendor-advisoryWEB
- nvd.nist.gov/vuln/detail/CVE-2025-65995ghsaADVISORY
- www.openwall.com/lists/oss-security/2025/12/12/2ghsaWEB
News mentions
0No linked articles in our index yet.