Summary
Open WebUI contains a server-side request forgery vulnerability in its web retrieval endpoints. An authenticated user may cause an Azure-hosted instance to request and return content from the Azure platform channel or other reserved network ranges when address validation incorrectly treats them as external.
Affected context
The issue occurs in Open WebUI deployments that expose the web retrieval or web search functionality and rely on Python's globally routable address classification to determine whether a destination is external. Azure-hosted instances are particularly relevant because the platform channel address may remain reachable despite this classification, although other reserved ranges may also be affected.
Exploit preconditions
- Attack surface: Web retrieval and web search API endpoints
- Required access: An authenticated Open WebUI account with permission to invoke the affected functionality
- Trigger point: Processing a user-controlled web destination through the retrieval service
- Why exploitation is possible: Address validation relies on a classification that does not reject every reserved or platform-specific destination, allowing server-side requests to reach internal resources
Why this happens
The underlying bug is insufficient validation of user-controlled network destinations. The implementation treated Python's globally routable classification as conclusive evidence that a destination was external, even though some reserved and cloud-provider platform addresses were not excluded. This maps to CWE-918: Server-Side Request Forgery, where an application makes outbound requests on behalf of a user without adequately restricting reachable destinations.
Potential impact
- An attacker may retrieve content from cloud platform channels
- Sensitive metadata or internal service responses could be exposed
- Internal network resources may be probed through the application
- Cloud credentials or configuration data could be disclosed if reachable
- The instance could be used as a foothold for further internal reconnaissance
General mitigation guidance
Mitigation generally focuses on enforcing an explicit outbound request policy rather than relying on a single routability classification.
Common defensive patterns include:
- Resolving and validating destinations against denylisted private, loopback, link-local, reserved, multicast, and cloud-provider platform ranges
- Rechecking the resolved destination immediately before connection and following redirects safely
- Allowing outbound requests only to approved schemes, ports, hosts, or network zones
- Routing retrieval traffic through an isolated egress proxy with filtering controls
- Applying least-privilege network policies and limiting access to instance metadata services
- Returning only the minimum necessary response content and avoiding sensitive diagnostic details
How teams detect issues like this
Organizations can detect SSRF risks through:
- Static analysis rules for server-side HTTP clients that consume user-controlled URLs
- Code review checks for incomplete private-address and redirect validation
- Tests covering loopback, link-local, private, reserved, IPv6, encoded, and cloud-platform destinations
- Egress firewall, proxy, and DNS monitoring for unexpected requests to internal or metadata ranges
- Policy enforcement requiring centralized URL validation and approved outbound network paths
- Neutral automated security review tools that identify tainted URL flows and unsafe network access patterns

