Executive brief
Apache Airflow is an open-source platform used to schedule and monitor complex workflows. A security flaw in the log server allows an authenticated user with access to one workflow (DAG) to potentially view the logs of other workflows they are not authorized to see. This could lead to the exposure of sensitive task outputs, error traces, and business data in shared or multi-team environments.
Technical details
The vulnerability stems from the use of Python's `str.lstrip()` instead of `str.removeprefix()` when validating JWT 'sub' claims against requested log paths. Because `lstrip()` removes any character in a provided set rather than a literal string, an attacker with a valid JWT for one Dag can access logs for any other Dag whose name starts with a subset of the characters in the original Dag's ID. This allows an authenticated worker to bypass per-Dag isolation boundaries to enumerate and read logs, potentially leaking sensitive task data. The issue is fixed in version 3.2.2 by correctly extracting the path segment.
Affected products
- Apache Airflow >= 3.0.0, < 3.2.2
Timeline
- 2026-05-12: patched: Pull request merged into main branch
- 2026-05-31: disclosed: Initial disclosure on oss-security mailing list
- 2026-06-01: advisory: GitHub Advisory and NVD entry published