CVE-2025-27018: Apache Airflow MySQL Provider SQLi Flaw

CVE-2025-27018 Overview

CVE-2025-27018 is a SQL Injection vulnerability affecting the Apache Airflow MySQL Provider. The vulnerability exists in the dump_sql and load_sql functions, where user-supplied table parameters from the UI are not properly sanitized before being used in SQL queries. This allows authenticated attackers with DAG triggering capabilities to inject malicious SQL commands, potentially leading to unauthorized data access, data corruption, or modification of database contents.

Critical Impact

Authenticated users can exploit this SQL injection vulnerability through DAG parameters to execute arbitrary SQL commands, potentially compromising data integrity, confidentiality, and availability of MySQL databases connected to Apache Airflow.

Affected Products

• Apache Airflow MySQL Provider versions prior to 6.2.0
• Apache Airflow deployments using the MySQL Provider with DAGs utilizing dump_sql or load_sql functions
• Systems exposing DAG triggering capabilities to users

Discovery Timeline

• March 19, 2025 - CVE-2025-27018 published to NVD
• June 3, 2025 - Last updated in NVD database

Technical Details for CVE-2025-27018

Vulnerability Analysis

This SQL injection vulnerability arises from improper neutralization of special elements in user-controlled input within the Apache Airflow MySQL Provider. When users trigger a DAG that utilizes the dump_sql or load_sql functions, they can supply a table parameter through the Airflow UI. This parameter is passed directly into SQL queries without adequate input validation or parameterization, creating an injection point.

The vulnerability requires authentication and the ability to trigger DAGs, which limits the attack surface to users who already have some level of access to the Airflow system. However, within that context, a malicious user could craft table parameter values containing SQL metacharacters and additional SQL statements to manipulate database operations beyond the intended scope.

Root Cause

The root cause of CVE-2025-27018 is insufficient input sanitization in the MySQL Provider's dump_sql and load_sql functions. The table parameter passed from the UI is incorporated into SQL statements without proper escaping, parameterized queries, or allowlist validation. This violates secure coding practices for database interactions where user input should never be directly concatenated into SQL query strings.

Attack Vector

The attack leverages the network-accessible Airflow web interface where authenticated users can trigger DAGs with custom parameters. An attacker with low-level privileges can:

1. Access the Airflow UI with valid credentials
2. Identify DAGs that use the vulnerable dump_sql or load_sql functions
3. Trigger the DAG execution with a maliciously crafted table parameter containing SQL injection payloads
4. Execute unauthorized SQL commands against the connected MySQL database

The attack exploits improper input handling where user-provided table names are directly interpolated into SQL queries. An attacker could craft a table parameter value containing SQL syntax that breaks out of the intended query structure and executes arbitrary commands. For example, a malicious table name like users; DROP TABLE sensitive_data; -- could be used to manipulate database operations beyond the intended scope.

For technical implementation details, see GitHub Pull Request #47254 and GitHub Pull Request #47255 which contain the security fixes.

Detection Methods for CVE-2025-27018

Indicators of Compromise

• Unusual SQL query patterns in MySQL logs containing union selects, subqueries, or comment sequences originating from Airflow connections
• DAG execution logs showing table parameters with SQL metacharacters such as semicolons, quotes, or comment delimiters
• Unexpected database modifications, deletions, or data exfiltration activities correlated with Airflow DAG triggers
• Failed authentication or authorization errors in MySQL following Airflow DAG executions

Detection Strategies

• Implement database activity monitoring to detect anomalous SQL queries executed through Airflow MySQL connections
• Review Airflow DAG trigger logs for suspicious table parameter values containing SQL injection patterns
• Deploy web application firewalls (WAF) with SQL injection detection rules on the Airflow web interface
• Enable MySQL general query logging temporarily to audit all queries executed by the Airflow service account

Monitoring Recommendations

• Configure alerting on Airflow DAG execution events involving MySQL Provider functions dump_sql and load_sql
• Monitor MySQL error logs for syntax errors that may indicate failed injection attempts
• Set up anomaly detection for database schema changes or bulk data operations
• Track and audit user activities in Airflow UI, particularly DAG triggering with custom parameters

How to Mitigate CVE-2025-27018

Immediate Actions Required

• Upgrade Apache Airflow MySQL Provider to version 6.2.0 or later immediately
• Audit existing DAGs using dump_sql or load_sql functions and restrict access to users who absolutely require it
• Review database user permissions for Airflow service accounts and apply least-privilege principles
• Enable additional logging on MySQL databases accessed by Airflow to detect potential exploitation attempts

Patch Information

Apache has released version 6.2.0 of the Airflow MySQL Provider which addresses this SQL injection vulnerability. The fix implements proper input sanitization for the table parameter in dump_sql and load_sql functions. Users should upgrade to this version as soon as possible.

The security patches are documented in GitHub Pull Request #47254 and GitHub Pull Request #47255. For additional information, refer to the Apache Mailing List Discussion.

Workarounds

• Restrict DAG triggering permissions to trusted administrative users only until the upgrade can be completed
• Implement input validation at the DAG level to allowlist only expected table names before passing to MySQL Provider functions
• Deploy network segmentation to limit direct database access from Airflow workers
• Consider temporarily disabling DAGs that utilize the vulnerable dump_sql or load_sql functions until patched

# Upgrade Apache Airflow MySQL Provider to patched version
pip install apache-airflow-providers-mysql>=6.2.0

# Verify installed version
pip show apache-airflow-providers-mysql | grep Version