OperTraitor — Kubernetes Operator RBAC Privilege-Escalation Scanner

What it is

OperTraitor is a Palo Alto Networks open-source scanner that audits RBAC manifests of Kubernetes operators — both locally installed ones and the full OperatorHub catalog — using an LLM to compare each operator's stated purpose against its actual granted permissions.

How it works

The tool parses RBAC YAML (Roles, ClusterRoles, ServiceAccounts) and feeds the operator's declared function alongside its permission set to an LLM-powered analysis engine, which flags mismatches and assigns a normalized 1-10 risk score. The underlying risk pattern: operators often get wildcard or cluster-wide ClusterRole permissions for convenience rather than namespace-scoped Roles; if an attacker compromises the operator (vulnerable image, dependency flaw, supply-chain attack), those excess permissions turn a contained breach into cluster-wide compromise.

Preconditions / Affected

Particularly relevant for older or abandoned operators still distributed via OperatorHub and the Operator Lifecycle Manager (OLM) — vendors may have shipped fixes via Helm/GitHub/ArtifactHub while the stale, over-privileged version remains the one pulled from default registries. Also relevant to the emerging class of LLM-enhanced/agentic operators that make autonomous decisions and call external services, where excess RBAC compounds into unintended autonomous actions, not just data exposure.

What to do

• Review every operator's service account permissions.
• Avoid deploying outdated packages from default registries; check for newer Helm/GitHub/ArtifactHub releases.
• Replace cluster-wide ClusterRoles with namespace-scoped Roles wherever the operator's actual function allows.
• Monitor Kubernetes audit logs for anomalous behavior, e.g., an operator reading Secrets from namespaces it has no functional reason to touch.

Caveats

Datadog disputed that its flagged permissions were excessive, stating some broad Secrets/RBAC access is required because users can define secret names dynamically, making them hard to restrict in advance; Datadog published documentation and mitigations instead of narrowing the permission set, so this finding remains a vendor-contested case rather than a confirmed vulnerability like the IBM one.