A server-side request forgery vulnerability in labring FastGPT versions 4.14.11 and prior allows authenticated attackers to bypass cloud metadata restrictions, posing a high risk of sensitive data exposure.

This is a Server-Side Request Forgery (CWE-918) vulnerability located in the isInternalAddress function within packages/service/common/system/utils.ts, where low-privileged authenticated attackers can bypass hardcoded blocklist checks using URL encoding techniques.

A successful exploit allows attackers to query internal cloud metadata endpoints and potentially access sensitive configuration data or credentials stored within the hosting environment. This can lead to unauthorized data exposure and further compromise of cloud infrastructure. Although the CVSS score is 7.7, the potential for unauthorized access to cloud control planes warrants prompt defensive action.

Immediate Action: Review the vendor security advisory and restrict outbound network traffic from the FastGPT application servers to sensitive metadata endpoints.

Proactive Monitoring: Monitor application logs for unusual outbound requests originating from the FastGPT container or server instance toward cloud metadata IP addresses.

Compensating Controls: Implement strict egress filtering and use Web Application Firewalls to inspect and block requests containing suspicious URL encoding patterns intended for internal IP ranges.

Public Exploit Available: No (false)

Security teams must treat this vulnerability with high priority despite the absence of a public patch at the time of publication. Administrators should enforce egress network controls and monitor for unauthorized internal requests while awaiting an official software update from the vendor.