CVE-2024-45034: Apache Airflow RCE Vulnerability

CVE-2024-45034 Overview

A privilege escalation vulnerability exists in Apache Airflow versions before 2.10.1 that allows DAG authors to execute arbitrary code on the scheduler. The vulnerability stems from improper handling of local settings files within DAG folders, enabling malicious DAG authors to add custom local settings that get executed by the scheduler process. This represents a significant security boundary violation, as the scheduler is not designed to execute code submitted by DAG authors.

Critical Impact
DAG authors can achieve code execution on the scheduler component, potentially compromising the entire Airflow deployment and gaining access to sensitive credentials, database connections, and other DAGs.

Affected Products

Apache Airflow versions prior to 2.10.1
Apache Airflow scheduler component
Apache Airflow DAG processing functionality

Discovery Timeline

2024-09-07 - CVE-2024-45034 published to NVD
2025-06-03 - Last updated in NVD database

Technical Details for CVE-2024-45034

Vulnerability Analysis

This vulnerability represents a privilege escalation through code injection in Apache Airflow's DAG processing mechanism. The core issue lies in how the Airflow scheduler handles local settings files within DAG folders. Under normal operation, the scheduler should only parse and schedule DAG definitions without executing arbitrary code from DAG authors. However, due to insufficient isolation between DAG content and scheduler execution context, a malicious DAG author can inject local settings files that the scheduler will load and execute with scheduler-level privileges.

The attack is particularly concerning because it breaks the intended security model where DAG authors are considered less privileged than scheduler operators. This vulnerability allows horizontal privilege escalation from a DAG authoring role to scheduler execution privileges.

Root Cause

The root cause is classified under CWE-250 (Execution with Unnecessary Privileges). The Airflow scheduler processes DAG folders in a manner that allows local Python settings files to be loaded and executed. The scheduler fails to properly sandbox or restrict the execution of code originating from DAG folders, treating local settings as trusted code when they should be considered untrusted user input.

The fundamental design flaw is the lack of separation between the DAG parsing context and the scheduler's privileged execution environment. When the scheduler scans DAG folders for workflow definitions, it inadvertently processes local settings files with its full privilege level.

Attack Vector

An attacker with DAG authoring privileges can exploit this vulnerability through the following approach:

The attacker creates or modifies a local settings file within a DAG folder
When the scheduler processes the DAG folder, it loads the malicious local settings
The code within the local settings executes with scheduler privileges
The attacker can then access sensitive configuration, credentials, or execute arbitrary commands

The attack requires network access and low-privilege authentication (DAG authoring permissions), but does not require user interaction. The vulnerability allows for complete compromise of confidentiality, integrity, and availability of the affected system.

Since this is a configuration and code injection vulnerability, the exploitation mechanism involves placing Python code in local settings files. Detailed technical information about the fix can be found in the GitHub Pull Request that addresses this vulnerability.

Detection Methods for CVE-2024-45034

Indicators of Compromise

Unexpected or unauthorized files appearing in DAG folders, particularly Python settings files
Unusual process spawning from the Airflow scheduler process
Modifications to local settings files that were not authorized by administrators
Anomalous network connections or system calls originating from the scheduler

Detection Strategies

Monitor DAG folder contents for new or modified files, especially any local settings or configuration files
Implement file integrity monitoring on DAG directories to detect unauthorized changes
Review scheduler logs for unusual execution patterns or errors related to local settings loading
Audit DAG author activities and correlate with scheduler execution events

Monitoring Recommendations

Enable verbose logging on the Airflow scheduler to capture all file loading operations
Deploy endpoint detection solutions to monitor scheduler process behavior for code execution anomalies
Implement alerting on changes to DAG folder structure that occur outside of approved deployment pipelines
Monitor for privilege escalation attempts or unusual credential access patterns from the scheduler context

How to Mitigate CVE-2024-45034

Immediate Actions Required

Upgrade Apache Airflow to version 2.10.1 or later immediately
Audit all existing DAG folders for unauthorized local settings files
Review and restrict DAG authoring privileges to trusted personnel only
Implement strict change management controls for DAG folder contents

Patch Information

Apache has released version 2.10.1 which fixes this vulnerability. The patch addresses the improper handling of local settings in DAG folders by preventing the scheduler from executing code submitted by DAG authors. Users should upgrade to this version or later as soon as possible.

For detailed information about the fix, refer to the Apache Mailing List Thread and the GitHub Pull Request. Additional security context is available in the Openwall OSS Security Update.

Workarounds

Restrict filesystem write access to DAG folders to only trusted administrators, not DAG authors
Implement a code review process for all DAG submissions before deployment to production
Use containerization or sandboxing to isolate the scheduler from sensitive system resources
Deploy Airflow in a configuration where DAG folders are read-only to the scheduler process

bash

# Configuration example: Restrict DAG folder permissions
chmod 755 /path/to/dags
chown airflow-admin:airflow /path/to/dags
# Remove write access for DAG authors at filesystem level
setfacl -m g:dag-authors:r-x /path/to/dags