CVE-2025-66236: 
Apache Airflow vulnerability analysis and mitigation

CVE-2025-66236 is a vulnerability in Apache Airflow where secrets from the Airflow configuration file are logged in plain text in the DAG run logs UI. The root cause is that the deprecated import path airflow.sdk.execution_time.secrets_masker redirected mask_secret() to a single-process implementation that only masked secrets in the task subprocess, leaving the supervisor process (which writes logs) unaware of masking instructions. This affects Apache Airflow versions 3.0.0 through 3.2.0 (exclusive). It was disclosed on April 13, 2026, with a CVSS v3.1 base score of 7.5 (High) per Feedly/NVD and a CVSS v4 score of 5.3 (Moderate) per the GitHub Advisory (GitHub Advisory, Openwall OSS-Security).

The vulnerability is classified as CWE-532 (Insertion of Sensitive Information into Log File). When DAG authors use the deprecated import path airflow.sdk.execution_time.secrets_masker to call mask_secret(), the compatibility shim previously redirected to a single-process implementation that only masked secrets within the task subprocess. Because the supervisor process — which is responsible for writing task logs — did not receive masking instructions via the inter-process messaging channel, secrets appeared in plain text in the DAG run logs UI visible to users with log access. The fix, implemented in PR #58662, redirects mask_secret() through the supervisor-aware implementation in airflow.sdk.log so that both the task and supervisor processes receive masking instructions (GitHub PR #58662, Openwall OSS-Security).

Successful exploitation results in sensitive secrets (such as credentials, API keys, or tokens stored in the Airflow config file) being exposed in plain text within DAG run task logs visible through the Airflow UI. Any user with access to view task logs — including lower-privileged Airflow users — could read these secrets, potentially enabling lateral movement, credential theft, or unauthorized access to downstream systems integrated with Airflow. There is no integrity or availability impact; the vulnerability is limited to confidentiality (GitHub Advisory, Openwall OSS-Security).

There is no public proof-of-concept exploit and no evidence of in-the-wild exploitation at this time. The EPSS score is approximately 0.042–0.095%, indicating a low probability of exploitation in the near term. The vulnerability is not listed in the CISA Known Exploited Vulnerabilities (KEV) catalog. Exploitation requires an attacker to already have access to the Airflow UI with permissions to view task logs, limiting the attack surface to authenticated users (GitHub Advisory, Openwall OSS-Security).

1. Gain Airflow UI access: Obtain credentials for an Airflow account with at least read access to DAG run logs (e.g., a Viewer or Operator role).
2. Identify vulnerable DAGs: Look for DAGs that import mask_secret from the deprecated path airflow.sdk.execution_time.secrets_masker and call it with sensitive values.
3. Trigger a DAG run: Execute or wait for a DAG run that invokes mask_secret() via the deprecated import path, causing the supervisor process to log secrets unmasked.
4. Access task logs: Navigate to the DAG run logs in the Airflow UI and retrieve the plain-text secrets exposed in the log output (GitHub PR #58662, Openwall OSS-Security).

• Logs: DAG run task logs in the Airflow UI containing plain-text secrets, credentials, API keys, or tokens that should have been masked (appearing as literal strings rather than ***).
• File System: Airflow task log files on disk (typically under the configured base_log_folder) containing unmasked sensitive values from the Airflow config.
• Code: DAG files importing mask_secret from airflow.sdk.execution_time.secrets_masker (the deprecated path) rather than from airflow.sdk.log.

Upgrade to Apache Airflow 3.2.0 or later, which includes the fix that properly redirects mask_secret() through the supervisor-aware implementation. As an interim measure, DAG authors should update any imports of mask_secret from the deprecated path airflow.sdk.execution_time.secrets_masker to use airflow.sdk.log directly. Deployment Managers should also review the updated Airflow security model documentation, workload isolation guidelines, and JWT authentication details to ensure their deployments are properly secured (GitHub Advisory, Openwall OSS-Security).

The vulnerability was reported by Saurabh Banawar and remediated by Amogh Desai of the Apache Airflow project. The disclosure was made via the oss-security mailing list on April 13, 2026, and accompanied by the Airflow 3.2.0 blog announcement highlighting broader security improvements. Community reaction has been limited, with the issue noted on security tracking platforms such as VulDB and CVEFeed (Openwall OSS-Security, GitHub Advisory).