/

CVE-2026-45726

Severity:

HIGH RISK

(7.6)

Sensitive Data Exposure (CWE-200) in Omni Resources

CVE-2026-45726

Deferred

Vulnerability type

Category:

Sensitive Data Exposure

CWE:

CWE-200, CWE-522, CWE-732

Impact:

Exposure of cryptographic secrets and unauthorized cluster control

Severity:

HIGH RISK

(7.6)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

Omni can expose the complete certificate-authority secrets bundle for an imported Talos cluster through an ImportedClusterSecrets resource. An authenticated user with the Reader role may retrieve unrotated Kubernetes, Talos, and etcd private keys, potentially enabling privileged cluster access beyond Omni's authorization boundary.

Affected context

The issue occurs when a standalone Talos cluster is imported into Omni and its CA secrets remain unrotated by the importing actor. In affected deployment paths, the resulting ImportedClusterSecrets resource can be retrieved through ResourceService by an authenticated Reader, exposing private keys associated with Kubernetes, Talos, etcd, and service-account authentication.

Exploit preconditions

  • Attack surface: Omni's ResourceService and ImportedClusterSecrets resource
  • Required access: An authenticated Omni account with the Reader role
  • Trigger point: Retrieval of the imported cluster's secrets resource
  • Required state: The importing actor has not rotated the cluster secrets
  • Why exploitation is possible: Access rules permit a Reader to retrieve sensitive private-key material outside the intended authorization boundary

Why this happens

The underlying issue is insufficient access control and protection for highly sensitive cryptographic material. A resource containing complete cluster CA secrets is exposed to a role that should not necessarily be able to read private keys, mapping primarily to CWE-200: Exposure of Sensitive Information. The disclosure is also related to CWE-522: Insufficiently Protected Credentials and CWE-732: Incorrect Permission Assignment for Critical Resource, because private keys are treated as accessible resource data despite their ability to authenticate or sign privileged identities.

Potential impact

  • An attacker may obtain Kubernetes, Talos, and etcd CA private keys.
  • An attacker could sign certificates for privileged Kubernetes identities such as system:masters.
  • An attacker could gain control of the imported cluster outside Omni's authorization boundary.
  • An attacker may access or modify workloads, credentials, and secrets.
  • The exposed service-account key could enable additional unauthorized authentication.

General mitigation guidance

Mitigation generally focuses on keeping private key material outside ordinary resource retrieval paths and enforcing least-privilege access controls.

Common defensive patterns include:

  • Preventing CA bundles and private keys from being stored in broadly readable application resources.
  • Applying explicit deny-by-default authorization to cryptographic secrets and credential-bearing resources.
  • Separating metadata access from secret retrieval and requiring narrowly scoped administrative authorization for sensitive operations.
  • Rotating or invalidating exposed cluster and service-account credentials when disclosure is suspected.
  • Encrypting sensitive secrets at rest and restricting backend, audit, and operational access.
  • Designing imports so secret handling does not extend the management platform's authorization boundary.

How teams detect issues like this

Organizations can look for:

  • Static analysis and policy checks for resources that include private keys, CA bundles, or service-account credentials.
  • Code review signals where Reader-level permissions are applied to credential-bearing resources.
  • Authorization tests verifying that non-administrative roles cannot retrieve secret-bearing resources.
  • Audit-log searches for ImportedClusterSecrets retrievals by Reader accounts.
  • Secret-scanning controls across API responses, resource definitions, logs, and persistent storage.
  • Automated security review tools that flag sensitive data exposure and excessive resource permissions.

Sources

SourceLink
github.comgithub.com/siderolabs/omni/commit/25fa9e141ee00285e70210ed40d11e4e457c871d
github.comgithub.com/siderolabs/omni/commit/a224cb020f3aaca2bc21c3e32a38eb4a37e531e8
github.comgithub.com/siderolabs/omni/commit/b8ca100c4539ea83f0ac2e53dcd523e6e5b68111
github.comgithub.com/siderolabs/omni/pull/2807
github.comgithub.com/siderolabs/omni/releases/tag/v1.6.6
github.comgithub.com/siderolabs/omni/releases/tag/v1.7.3
github.comgithub.com/siderolabs/omni/security/advisories/GHSA-wv8c-6mx2-xf4j

Frequently asked

questions

It can be exploitable through the Omni service by an authenticated Reader, provided the imported cluster's secrets have not been rotated and the affected resource remains retrievable.

The key prerequisite is an authenticated Omni account with the Reader role. No direct access to the underlying Kubernetes cluster is necessarily required to retrieve the exposed resource.

Keep private keys out of ordinary resource APIs, enforce deny-by-default authorization for secret-bearing resources, and require narrowly scoped administrative access for any necessary secret operations.

Classify cryptographic material as highly sensitive, test role-based access controls, scan API responses and storage for private keys, and rotate credentials whenever disclosure may have occurred.

Find out what's already
exploitable in your codebase.