This vulnerability arises from the Data Science Pipelines Operator's insufficient validation of the spec.database.customExtraParams field, which is used to construct the MySQL Data Source Name (DSN). An attacker with namespace editor privileges can inject parameters that enable the LOCAL INFILE functionality in MySQL, facilitating the exfiltration of sensitive files from the operator pod, including the service account token. This bypasses namespace boundaries and allows escalation to cluster-admin privileges. The vulnerability is tracked as CVE-2026-18617 with a CVSS 3.1 score of 8.8 (high severity). Red Hat's advisory confirms the issue but states that no suitable mitigation or fix currently exists that meets their criteria for ease of use, applicability, and stability.

Successful exploitation allows a namespace editor to exfiltrate sensitive files from the operator pod, including service account tokens, by injecting malicious parameters into the MySQL DSN. This leads to privilege escalation, granting cluster-admin access and effectively bypassing Kubernetes namespace isolation. The impact includes full confidentiality, integrity, and availability compromise within the cluster scope.

As per the Red Hat advisory, no official fix or mitigation currently meets the criteria for deployment and stability. Users should monitor the Red Hat advisory for updates on patches or mitigations. Until a fix is available, restricting namespace editor privileges and limiting access to the Data Science Pipelines Operator may reduce exposure. Engage with Red Hat support or Technical Account Managers for tailored guidance.