| Title | GitHub Helicone ai-gateway v0.2.0-beta.30 Server-Side Request Forgery |
|---|
| Description | Helicone AI Gateway is vulnerable to Server-Side Request Forgery (SSRF) through improper handling of provider base URLs. The gateway allows providers to be configured with custom base-url values (defined in providers.yaml or user-supplied config.yaml). When processing requests, the application forwards LLM API calls to the configured upstream endpoint using these base URLs without sufficient validation or restrictions on the target host, scheme, or IP address range. An attacker capable of influencing or injecting a malicious provider base URL can coerce the gateway into making arbitrary HTTP requests to internal network services, cloud instance metadata endpoints (e.g., x.x.x.x), localhost services, or other restricted resources. This occurs because outbound requests are constructed directly from the attacker-controlled base URL without implementing allowlists, private IP blocking, or safe URL parsing. Successful exploitation may enable reconnaissance of internal infrastructure, access to sensitive services within the gateway’s network context, or further lateral movement. |
|---|
| Source | ⚠️ https://github.com/oscar2744/CVE-Submit/issues/1 |
|---|
| User | havel8964 (UID 98694) |
|---|
| Submission | 06/02/2026 05:32 (2 months ago) |
|---|
| Moderation | 07/12/2026 07:52 (1 month later) |
|---|
| Status | Accepted |
|---|
| VulDB entry | 377837 [Helicone ai-gateway up to 0.2.0-beta.30 AWS Metadata Service service.rs build_target_url extracted_path_and_query server-side request forgery] |
|---|
| Points | 20 |
|---|