OpenRemote Incomplete Fix for XXE in KNXProtocol Leads to Arbitrary File Read (CVE-2026-54640)

OpenRemote, an open-source IoT platform, has an incomplete fix for a previously identified XML External Entity (XXE) vulnerability, CVE-2026-40882. The original patch for CVE-2026-40882 only secured the Velbus asset import handler, leaving the KNXProtocol handler vulnerable to arbitrary file read. Versions up to and including 1.24.1 of the openremote-agent module are affected by this newly identified vulnerability, tracked as CVE-2026-54640. An authenticated, non-administrative user can exploit this flaw by uploading a specially crafted ETS project ZIP file containing malicious XML. This allows attackers to read arbitrary files from the server's filesystem, including /etc/passwd, application configuration files with credentials, or even leverage Server-Side Request Forgery (SSRF) to access internal network resources or cloud metadata endpoints. This poses a significant risk to the confidentiality of sensitive data and can lead to further compromise of the OpenRemote instance and its underlying infrastructure.

Attack Chain

1. An authenticated, non-administrative attacker sends an HTTP POST request to the /api/{realm}/agent/{agentId}/import endpoint with a malicious ETS project ZIP file as fileData.
2. The OpenRemote AgentResourceImpl.doProtocolAssetImport() method receives the fileData and passes it to KNXProtocol.startAssetImport().
3. KNXProtocol extracts the 0.xml file from the attacker-controlled ZIP archive using a ZipInputStream.
4. The content of 0.xml is processed by the Saxon TransformerFactoryImpl.newTransformer() and transform() methods without XXE protection, allowing external entities to be resolved.
5. Separately, XMLInputFactory.newInstance().createXMLStreamReader() also processes the 0.xml content without XXE protection.
6. Both XML parsers resolve an external entity defined in the malicious 0.xml (e.g., <!ENTITY xxe SYSTEM "file:///etc/passwd">), leading to arbitrary file read.
7. The contents of the arbitrary file are returned to the attacker within the response of the import process or are processed in a way that allows exfiltration.
8. Successful exploitation results in the attacker obtaining sensitive information from the server filesystem or probing internal network resources via SSRF.

Impact

This vulnerability allows any authenticated user, even without administrative privileges, to read arbitrary files from the server filesystem. This includes critical system files like /etc/passwd, enabling user enumeration and reconnaissance. More significantly, it allows access to application configuration files (e.g., openmrs-runtime.properties) that might contain sensitive database credentials, API keys, or other secrets. The potential for Server-Side Request Forgery (SSRF) further broadens the attack surface, allowing attackers to query cloud provider metadata endpoints (e.g., http://169.254.169.254/) for cloud credentials or access internal services reachable from the vulnerable OpenRemote server. The vulnerability is present in a core protocol handler (KNXProtocol) shipped with the OpenRemote agent module, making all default installations running affected versions susceptible to this critical data exfiltration risk.

Recommendation

• Patch CVE-2026-54640 immediately: Upgrade the io.openremote:openremote-agent module to a version patched against CVE-2026-54640.
• Review and audit configurations: Conduct an audit of all OpenRemote instances to identify any running openremote-agent modules and verify their versions, applying the necessary updates.
• Implement network segmentation: Restrict network access to cloud metadata endpoints and internal services from OpenRemote servers to mitigate the impact of potential SSRF exploitation, as mentioned in the Impact section.
• Enable webserver logging for API POST requests: Implement detailed logging for POST requests to /api/{realm}/agent/{agentId}/import on your web servers (e.g., NGINX, Apache, IIS). Monitor for unusually large or malformed XML payloads within the fileData parameter that might indicate attempted XXE exploitation, although this may require deep packet inspection or WAF capabilities.
• Apply WAF rules for XXE: Configure your Web Application Firewall (WAF) to detect and block XML External Entity (XXE) injection attempts in XML content submitted to the affected /api endpoints, specifically looking for <!DOCTYPE declarations containing SYSTEM or PUBLIC identifiers.