Summary
Label Studio through version 1.23.0 does not adequately validate webhook destinations, allowing authenticated users to direct server-side requests toward internal services, private-network addresses, or cloud metadata endpoints. When outbound payload transmission is enabled, an attacker may use these requests to access or exfiltrate annotation data from reachable services.
Affected context
The vulnerability occurs in Label Studio deployments where authenticated users can create or configure webhooks and the application can make outbound network requests on their behalf. Risk is higher when webhook destinations are not restricted to approved external endpoints, when private or metadata address ranges are reachable from the application environment, and when annotation data is included in webhook payloads.
Exploit preconditions
- Attack surface: Label Studio webhook configuration and outbound request handling
- Required access: An authenticated account with permission to create or modify webhooks
- Trigger point: Webhook delivery after an event causes Label Studio to send a request to the configured destination
- Why exploitation is possible: Destination validation does not reliably prevent requests to private networks, internal services, or cloud metadata endpoints
- Additional condition: Payload transmission must be enabled or otherwise configured to include annotation data
Why this happens
The underlying bug is insufficient validation and authorization of user-controlled webhook URLs. The application accepts destinations that may resolve to internal address ranges or privileged cloud endpoints and then performs server-side requests from a trusted network position. This behavior maps to CWE-918: Server-Side Request Forgery, because attacker-influenced URLs cause the server to access resources that may not be directly reachable by the attacker.
Potential impact
- An attacker may access internal services reachable from the Label Studio environment
- Cloud instance metadata or service credentials could be exposed
- Annotation data could be exfiltrated through outbound webhook requests
- Internal network topology or service information could be disclosed
- Compromised downstream services could be used for further lateral movement
General mitigation guidance
Mitigation generally focuses on reducing trust in user-supplied webhook destinations and enforcing outbound network boundaries.
Common defensive patterns include:
- Allowing webhooks only for explicitly approved destinations or domains
- Blocking loopback, link-local, RFC 1918, IPv6 private, and cloud metadata address ranges
- Resolving and validating destinations before connection, while accounting for DNS rebinding and redirects
- Restricting outbound traffic with network egress policies, firewalls, or service-mesh controls
- Avoiding sensitive annotation data in webhook payloads unless the destination is trusted
- Running the application with minimal network reachability and least-privilege cloud identity permissions
How teams detect issues like this
Detection approaches can include:
- Static analysis and code review checks for direct HTTP requests using user-controlled URLs
- Tests covering private, loopback, link-local, metadata, redirect, and DNS-rebinding destinations
- Monitoring webhook configuration changes and outbound requests to internal address ranges
- Egress policy alerts for connections from the application to metadata services or unexpected private networks
- Reviewing whether webhook payloads contain annotations or other sensitive information
- Automated security review tools can help identify missing URL allowlists and SSRF-prone request flows

