On this page (7 sections)

Incus, the container and virtual machine manager, has a critical flaw tracked as CVE-2026-130783. According to the NVD record, versions before 7.6.0 expose certificate and cluster join token secrets through the default project event stream, and also skip an authorization check when operations without associated permissions are cancelled.

If you run Incus servers or clusters, especially where some users or automation hold restricted client certificates, treat this as a priority upgrade. The NVD description says a restricted client can end up with full server access.

What happened

The NVD entry, published on 2026-11-18, describes two connected problems in how Incus handles operations. First, secrets such as certificate tokens and cluster join tokens appear in the event stream of the default project. Second, cancelling an operation that has no associated permissions skips the authorization step.

The record points to a GitHub security advisory from the Incus project (GHSA-cwg9-8mc4-hf7j). The research available for this article contains only the NVD description, so details beyond it, such as exact code paths or backported fixes, are not confirmed here. Read the advisory directly before planning your rollout.

How the weakness works, in plain terms

An event stream is a feed that tells a connected client what is happening on the server. Clients subscribe to it to watch operations such as image downloads, instance creation or token generation. The expectation is that a client only sees what it is permitted to see.

Here, the feed for the default project carried secret values that should have stayed private. A join or certificate token is effectively a one-time password for adding a new trusted identity. Anyone who reads it before the intended recipient can use it first. This is the classic information exposure weakness (CWE-200): the secret leaks, and the leak turns into an access problem.

The NVD description gives two outcomes:

  • A restricted client subscribes to operation events, captures a certificate token, and enrolls an unrestricted certificate. That gives it full server access.
  • A client cancels token operations by UUID, because the cancel path skips authorization when the operation has no associated permissions. This can disrupt legitimate enrollment or cluster joining.
The research does not state whether an attacker needs any existing credentials beyond a restricted client, or whether an administrator must be generating a token at the time. The wording points to a restricted but already authenticated client. Treat unauthenticated exposure as unconfirmed.

Who is realistically exposed

The highest risk is in setups where trust is deliberately split:

  • Shared Incus servers where some users hold restricted certificates limited to certain projects.
  • Clusters where administrators generate join tokens for new members.
  • Automation or CI systems that hold a restricted client certificate to the Incus API.
  • Hosting or lab environments that give tenants API access.

A single-user server where the only client is already a full administrator gains little from the token leak, since there is no privilege boundary to cross. Even so, upgrading is the right move, because the cancel authorization bug does not depend on how many users you have.

How to check your exposure

First find out which version each server runs. This read-only command prints client and server version information; run it on a machine with the Incus client configured for the server in question and look for a server version below 7.6.0.

Show Incus client and server versions (run from a configured client)Shell
incus version

Then list which certificates the server trusts. Run this as an administrator and look for any certificate you do not recognize, or any unrestricted certificate that was added at an unexpected time.

List trusted client certificates (run as an administrator)Shell
incus config trust list

Also review who currently holds restricted certificates and what they are used for. Any restricted client that existed while an unpatched server was running should be considered a possible source of a captured token.

What to do now

  1. Inventory every Incus server and cluster member and record its version with incus version.
  2. Upgrade to Incus 7.6.0 or later using your normal package or build process, and confirm the version afterwards.
  3. Hold off on generating new certificate or join tokens on unpatched servers, and avoid leaving unused tokens outstanding.
  4. Review the trust list with incus config trust list and remove any certificate you cannot account for.
  5. If a restricted client could have watched the event stream while a token was issued, treat that token as exposed and revoke or reissue related certificates.
  6. For clusters, check that the member list matches what you expect and that no unknown node has joined.
  7. Reduce API exposure by limiting network access to the Incus API to trusted management networks.
  8. Read the upstream advisory GHSA-cwg9-8mc4-hf7j and the Incus security advisories page for any additional guidance from the maintainers.
Caution. Upgrading a server can restart the Incus daemon and may interrupt API access during the change. Plan the upgrade in a maintenance window, and for clusters follow the upstream upgrade guidance so members stay compatible.

Removing a certificate you do not recognize is a judgment call. Confirm with your team that it is not used by a pipeline or colleague before revoking it, then replace it with a correctly scoped certificate if needed.

Detection and monitoring

The research gives no indicators of compromise, so there is no confirmed signature to search for. Practical signs worth checking are behavioral:

  • New entries in the trust store that were not requested by an administrator.
  • Cluster members that you did not add.
  • Token or enrollment operations that were cancelled or failed unexpectedly, which may point to interference.
  • Unfamiliar client addresses connecting to the Incus API.

Keep API access logs and server logs long enough to look back across the period an unpatched version was running. If you expose any service fronted by Incus through a web address, a monitoring service such as Webforta Monitor can alert you to certificate and DNS changes, which helps you notice unexpected changes around your infrastructure, though it does not see inside the Incus API itself.

For general hardening of exposed services, our guide to small business website security covers the basics of limiting access and reviewing who can reach your systems.

What is still unknown

  • Whether the flaw is being exploited in the wild. No source in the research says so, and no CISA KEV listing is shown.
  • Whether public proof-of-concept code exists. The research does not say.
  • Whether fixes were backported to older release lines. Only 7.6.0 is named.

Only one independent publisher, the NVD record, backs the technical description at this point, so details may be refined as the maintainers' advisory and other analysts add information.

Sources

This article is based on the public sources below. Details can change as vendors and researchers publish more; check the original advisories before acting on version numbers.

  1. CVE-2026-130783: Incus before 7.6.0 exposes certificate and cluster join token secrets through the default project event stream and skips authorization when cancelling…nvd.nist.gov ·
  • Incus
  • CVE-2026-130783
  • containers
  • information disclosure
  • cluster security