Apache Airflow Arbitrary File Read Vulnerability via Log Symlink in FileTaskHandler

A vulnerability exists in Apache Airflow versions prior to 3.2.2, allowing a DAG author to create a symlink in their task's log directory that points to an arbitrary file readable by the API server process. This read-path attack could target files like '/etc/passwd' or 'airflow.cfg'. Alternatively, the author could supply a 'task_id' with '..' sequences that bypass the Task SDK's 'KEY_REGEX', leading to a write-path attack. In both scenarios, the FileTaskHandler resolves the log path outside the designated 'base_log_folder', causing unauthorized leakage or overwriting of files. This issue arises only in deployments where the worker log folder is shared with the API server.

Impact

Exploitation of this vulnerability could result in unauthorized reading or modification of files on the API server's filesystem, depending on the nature of the symlink or the file overwritten.

Remediation

Users are advised to upgrade to Apache Airflow version 3.2.2 or later. As an additional precaution, deploy the worker and API server with separate log volumes to prevent worker-controlled paths from accessing the API server's filesystem.