Executive brief
Kyverno is a policy engine for Kubernetes that validates and controls container deployments. A vulnerability in its apiCall feature allows authenticated users to send arbitrary HTTP requests to internal services and cloud metadata endpoints, potentially exposing sensitive credentials and configuration data. Response data is reflected back in error messages, making data theft straightforward for attackers with cluster access.
Technical details
A server-side request forgery (SSRF) vulnerability exists in Kyverno's apiCall.service.url when used with variable substitution. The vulnerability stems from lack of URL validation and IP filtering in the HTTP client, allowing authenticated users to inject attacker-controlled input via resource annotations or other metadata fields. Requests can target loopback addresses (127.0.0.1), link-local addresses (169.254.0.0/16), AWS IMDS, and internal ClusterIP services. The HTTP client follows up to 10 redirects without re-validation, and response data from non-2xx responses or JSON parsing failures is reflected in admission error messages, enabling non-blind data exfiltration. Authentication is required (cluster user), but attack complexity is low. Patch is available in version 1.18.0.
Affected products
- Kyverno Kyverno before 1.18.0
Timeline
- 2026-04-13: advisory: GitHub security advisory published
- 2026-09-01: disclosed: CVE-2026-84196 published on NVD
- 2026-04-13: patched: Patch available in version 1.18.0