Apache Linkis: Arbitrary File Read via Double URL Encoding Bypass

An attacker exploiting CVE-2025-29847 could gain unauthorized access to system files within the Apache Linkis environment. This could lead to the exposure of sensitive configuration data, credentials, or even code. The ability to read arbitrary files significantly expands the attack surface, potentially enabling further exploitation and lateral movement within the network. While the description doesn't explicitly mention remote code execution, access to configuration files could be leveraged to modify Linkis behavior or inject malicious code. The scope of impact is broad, affecting any Linkis deployment using the JDBC engine and data source functionality with improperly encoded URLs.

Organizations deploying Apache Linkis for data processing and analytics are at risk, particularly those relying on the JDBC engine for connecting to external data sources. Environments with legacy Linkis installations (versions 1.3.0 - 1.7.0) are especially vulnerable, as are those with complex JDBC configurations or inadequate input validation.

• java / server:

find /opt/linkis/logs -type f -name "*.log" | grep -i "JDBC URL contains multiple encoded characters"

• generic web:

curl -I <linkis_jdbc_url_endpoint> | grep -i "URL-encoded characters"

1. 2026-01-19Disclosure

   disclosure

1. Reserviert2025-03-12
2. Veröffentlicht2026-01-19
3. Geändert2026-02-03
4. EPSS aktualisiert2026-03-23

The primary mitigation for CVE-2025-29847 is to upgrade Apache Linkis to version 1.8.0 or later, which includes the necessary URL validation fix. If upgrading immediately is not feasible, a temporary workaround involves implementing stricter URL validation on the frontend. This should include checking for the presence of the '%' character in connection information and performing URL decoding before processing. Consider implementing a Web Application Firewall (WAF) rule to block requests with excessively encoded URLs. After upgrading, confirm the fix by attempting to connect with a URL containing multiple encoded characters and verifying that the access is denied.