CVE-2026-39961: Secret Exfiltration in Aiven Operator

The core of the vulnerability lies in the Aiven Operator's trust of user-supplied namespace values within the connInfoSecretSource specification. The operator's ServiceAccount possesses cluster-wide read/write access to secrets (via aiven-operator-role ClusterRole), enabling it to access sensitive data. An attacker can craft a malicious ClickhouseUser CRD that instructs the operator to read secrets from a victim's namespace and write them to a new secret within the attacker's own namespace. This effectively allows the attacker to steal production database credentials, API keys, and other sensitive tokens without requiring direct access to the victim's namespace. The blast radius is significant, as any namespace containing secrets is potentially at risk.

1. Reserved2026-04-07
2. Published2026-04-10
3. EPSS updated2026-04-17

The primary mitigation is to upgrade the Aiven Operator to version 0.37.0 or later, which includes the necessary validation to prevent the confused deputy attack. If an immediate upgrade is not feasible, consider implementing stricter Role-Based Access Control (RBAC) policies to limit the permissions of the aiven-operator ServiceAccount. Specifically, restrict its ability to read secrets from namespaces it doesn't explicitly need to access. Additionally, review and audit existing ClickhouseUser CRDs to identify and remove any potentially malicious configurations. After upgrading, confirm the fix by verifying that the operator no longer attempts to read secrets from unauthorized namespaces using Kubernetes audit logs.