An unauthenticated server-side request forgery vulnerability in Portkey AI Gateway allows attackers to exfiltrate internal data and API keys.

The application fails to apply the requestValidator middleware to the /v1/proxy/* route, allowing unauthenticated attackers to manipulate the x-portkey-custom-host header. This flaw enables the forwarding of requests to arbitrary internal network addresses, including those carrying sensitive Authorization headers.

Successful exploitation allows an attacker to bypass perimeter security to interact with internal services that are not intended to be publicly accessible. This can lead to the exfiltration of sensitive provider API keys, unauthorized access to internal infrastructure, and potential compromise of the broader production environment. With a CVSS score of 7.5, this high-severity vulnerability represents a significant risk to data confidentiality and internal system integrity.

Immediate Action: Upgrade Portkey AI Gateway to a version beyond 1.15.2 where the requestValidator middleware is correctly applied to all proxy routes.

Proactive Monitoring: Review web server and application access logs for unusual requests to the /v1/proxy/* endpoint, specifically monitoring for the presence of the x-portkey-custom-host header pointing to internal IP addresses or private hostnames.

Compensating Controls: Implement strict egress filtering on the host machine to prevent the gateway service from initiating unauthorized connections to internal network segments or unauthorized external destinations.

Public Exploit Available: Unknown

Given the capability to exfiltrate API keys and reach internal services, administrators must prioritize this update. Ensure all instances of the Portkey AI Gateway are patched to the latest version to close the proxy validation gap and prevent unauthorized internal network traversal.