/
← Voltar ao feed
medium

Unit 42 releases OperTraitor, finds over-privileged Kubernetes operators incl. IBM Prometurbo (CVE-2026-6389) and Datadog Operator

1 fonte(s) · 1 atualização(ões) · publicado 29/09/2026 07:00 · coletado há 1 dia

Unit 42 (Palo Alto Networks) released OperTraitor, an open-source, LLM-powered tool that ingests RBAC configurations from installed Kubernetes operators and the OperatorHub catalog. It compares an operator's documented function with its granted privileges and assigns a 1-10 risk score. The research found abandoned, overly permissive operators in OperatorHub, with slightly over 5% of the operators reviewed requesting excessive privileges, including implicit paths to cluster admin. Two case studies are described: IBM Turbonomic Prometurbo (CVE-2026-6389, CVSS 8.8, patched) and the Datadog Operator, where the broad permissions are a documented design trade-off. No in-the-wild exploitation is reported. The article also warns that agentic, LLM-driven operators will turn these RBAC misconfigurations into active threat vectors.

Technical Details

  • A compromised operator's blast radius is defined entirely by its service account's RBAC permissions, whether the compromise comes from a supply chain attack, a dependency vulnerability or node hijacking.
  • Prometurbo (IBM): the service account was bound to a ClusterRole granting get, list and watch on secrets in the core API group cluster-wide. If compromised, an attacker could dump service account tokens, database credentials, API keys and TLS certificates from unrelated namespaces. The OperatorHub version (v8.6.0, 2022) was outdated, and the finding was confirmed on the GitHub version v8.17.6.
  • Datadog Operator: cluster-wide access to secrets plus various verbs on ClusterRoles and ClusterRoleBindings. Datadog explained that secret names are user-defined and cannot be predicted, and it added documentation of its RBAC settings and applied mitigations.
  • Supply chain weakness: vendors publish newer, secure operator versions via Helm, GitHub or ArtifactHub, while older vulnerable versions stay available through the Operator Lifecycle Manager (OLM). Many operator owners did not respond to disclosure attempts.
  • Agentic risk patterns: LLM-enhanced logic (e.g. K8sGPT), external agent bridges (e.g. via Model Context Protocol) and full agent runtimes inheriting broad RBAC.
  • Disclosure timeline: reported to IBM on Nov. 5, 2025; fix confirmed on Feb. 3, 2026; IBM bulletin and CVE published on April 24, 2026.

Remediation

  • Do not implicitly trust operators from OLM or OperatorHub. Deploy the latest versions from maintained Helm charts, ArtifactHub or official GitHub repositories.
  • Prefer namespace-scoped operators and avoid ClusterRoles and ClusterRoleBindings unless necessary.
  • Validate vendor YAML manifests and continuously audit and downscope RBAC, for example with OperTraitor.
  • Enable Kubernetes Audit Logs, baseline operator service account behavior and alert on anomalies such as listing secrets in unrelated namespaces or API calls from unexpected IPs.
  • For LLM-enhanced or agentic operators, enforce strict network policies that block public internet and unauthorized internal access, and limit the context and permissions passed to the LLMs.
  • Apply the IBM patch for CVE-2026-6389.

Linha do tempo

  1. unit42.paloaltonetworks.com ↗ · pub. 29/09/2026 07:00 · coletado 19/10 17:56 primeiro registro

    Unit 42 releases OperTraitor, finds over-privileged Kubernetes operators incl. IBM Prometurbo (CVE-2026-6389) and Datadog Operator

    ler a análise desta fonte

    Unit 42 (Palo Alto Networks) released OperTraitor, an open-source, LLM-powered tool that ingests RBAC configurations from installed Kubernetes operators and the OperatorHub catalog. It compares an operator's documented function with its granted privileges and assigns a 1-10 risk score. The research found abandoned, overly permissive operators in OperatorHub, with slightly over 5% of the operators reviewed requesting excessive privileges, including implicit paths to cluster admin. Two case studies are described: IBM Turbonomic Prometurbo (CVE-2026-6389, CVSS 8.8, patched) and the Datadog Operator, where the broad permissions are a documented design trade-off. No in-the-wild exploitation is reported. The article also warns that agentic, LLM-driven operators will turn these RBAC misconfigurations into active threat vectors.

    Technical Details

    • A compromised operator's blast radius is defined entirely by its service account's RBAC permissions, whether the compromise comes from a supply chain attack, a dependency vulnerability or node hijacking.
    • Prometurbo (IBM): the service account was bound to a ClusterRole granting get, list and watch on secrets in the core API group cluster-wide. If compromised, an attacker could dump service account tokens, database credentials, API keys and TLS certificates from unrelated namespaces. The OperatorHub version (v8.6.0, 2022) was outdated, and the finding was confirmed on the GitHub version v8.17.6.
    • Datadog Operator: cluster-wide access to secrets plus various verbs on ClusterRoles and ClusterRoleBindings. Datadog explained that secret names are user-defined and cannot be predicted, and it added documentation of its RBAC settings and applied mitigations.
    • Supply chain weakness: vendors publish newer, secure operator versions via Helm, GitHub or ArtifactHub, while older vulnerable versions stay available through the Operator Lifecycle Manager (OLM). Many operator owners did not respond to disclosure attempts.
    • Agentic risk patterns: LLM-enhanced logic (e.g. K8sGPT), external agent bridges (e.g. via Model Context Protocol) and full agent runtimes inheriting broad RBAC.
    • Disclosure timeline: reported to IBM on Nov. 5, 2025; fix confirmed on Feb. 3, 2026; IBM bulletin and CVE published on April 24, 2026.

    Remediation

    • Do not implicitly trust operators from OLM or OperatorHub. Deploy the latest versions from maintained Helm charts, ArtifactHub or official GitHub repositories.
    • Prefer namespace-scoped operators and avoid ClusterRoles and ClusterRoleBindings unless necessary.
    • Validate vendor YAML manifests and continuously audit and downscope RBAC, for example with OperTraitor.
    • Enable Kubernetes Audit Logs, baseline operator service account behavior and alert on anomalies such as listing secrets in unrelated namespaces or API calls from unexpected IPs.
    • For LLM-enhanced or agentic operators, enforce strict network policies that block public internet and unauthorized internal access, and limit the context and permissions passed to the LLMs.
    • Apply the IBM patch for CVE-2026-6389.