Between November 10 and 11, 2026, VulnCheck (acting as CNA) filed three 9.8 CVEs in NVD against three open source projects you run yourself: the Kubero PaaS, the WWBN AVideo video platform and the NextScan vulnerability scanner. The outcome is the same in each case: an unauthenticated attacker logs in as an administrator, no password needed.
Two of the three share the same mistake: a hard-coded JWT signing secret. If the key sits in the source code or a public Docker image, anyone can mint a valid token.
| CVE | Product | Versions | Flaw | CVSS 3.1 (VulnCheck) | NVD status |
|---|---|---|---|---|---|
| CVE-2026-125370 | Kubero | 3.0.0 through 3.1.1 | Hard-coded fallback JWT secret (CWE-321) | 9.8 | Deferred |
| CVE-2026-126542 | WWBN AVideo | 6.5 through 29.7.1 | Impersonation via social login display name (CWE-290) | 9.8 | Deferred |
| CVE-2026-126797 | NextScan | 1.1.0 through 1.2.0 | Hard-coded JWT signing key (CWE-321) | 9.8 | Deferred |
VulnCheck also gives each one a CVSS 4.0 score of 9.3. All three carry the vector AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H. NVD hasn't analyzed them and attaches no CPE configuration. None is in the CISA KEV catalog at the time of writing.
Kubero: CVE-2026-125370
Kubero is a self-hosted PaaS built on Kubernetes. In versions 3.0.0 through 3.1.1, when the JWT_SECRET environment variable is not set, the server falls back to a hard-coded signing secret. An attacker who knows it signs a token with whatever claims they want: role, userGroups, permissions.
A second flaw makes it worse: JwtStrategy.validate() trusts those claims without a database lookup. The forged token doesn't even need to match a real user. The attacker then controls:
- pipelines and deployed applications;
- users;
- API tokens.
On a PaaS, controlling pipelines in practice means being able to ship code to the cluster. That's our inference; the NVD description stops at control over the objects listed above.
The immediate fix follows from the description: set JWT_SECRET to a random value unique to your installation. NVD names no fixed version; issue 786 on the kubero-dev/kubero repository, cited as a reference, is where to follow the fix.
WWBN AVideo: CVE-2026-126542
AVideo supports login through X, Facebook, Yahoo or LinkedIn accounts (via the Hybridauth library). From 6.5 through 29.7.1, when the provider doesn't return an email address, the app matches the social account to a local account by display name.
The attack:
- the attacker sets their social account's display name to the victim's username;
- they withhold their email during authorization;
login.json.phporoauth2.php(MobileManager plugin) logs them in as the victim, administrators included.
A social network display name is a free-text field anyone can change at will. It was never proof of identity. The NVD references point to GitHub advisory GHSA-9rvm-cfc4-q9g9 and a fix commit in the WWBN/AVideo repository. Check the advisory for the first release that includes it; the NVD description doesn't name one.
NextScan: CVE-2026-126797
NextScan is an open source vulnerability scanner. Its ns-admin component, versions 1.1.0 through 1.2.0, signs its tokens with a hard-coded JWT key. An attacker signs an HS256 X-Access-Token carrying the super administrator role ID, which is also public, and gets full access to:
- user management;
- scan tasks;
- stored vulnerability results.
That last one is the real problem. A scanner holds the list of weaknesses in your own network, so its results amount to a ready-made attack map. NVD cites the v1.2.0 docker-compose.yaml and the lysec/ns-admin Docker image, which suggests a deployment that follows the default setup is affected. NVD names no fixed version; follow issue 48 on the project repository.
Affected versions
| Product | Affected | Fixed version per NVD |
|---|---|---|
| Kubero | 3.0.0 through 3.1.1 | not stated (set JWT_SECRET) |
| WWBN AVideo | 6.5 through 29.7.1 | not stated, fix commit referenced |
| NextScan (ns-admin) | 1.1.0 through 1.2.0 | not stated |
These bounds come from the VulnCheck descriptions, not a NIST analysis. A version above the upper bound isn't automatically fixed: check each project's repository before drawing conclusions.
Detection
- Kubero: confirm
JWT_SECRETis actually set in your deployment. Look for unexplained API tokens or users, and pipeline changes outside planned work. - AVideo: logins through
login.json.phporoauth2.phpon accounts that don't normally use social login, especially admin accounts. - NextScan: requests carrying an
X-Access-Tokenfrom unfamiliar addresses, new accounts, or exports of scan results. - For all three: is the admin interface reachable from the internet?
Mitigation
1. Replace every default secret
Kubero: set JWT_SECRET. NextScan: apply the project's fix as soon as it ships and, until then, keep ns-admin off the network. Once the secret changes, every token signed with the old one stops working.
2. Turn off social login on AVideo
Until you've upgraded, disable the X, Facebook, Yahoo and LinkedIn providers, or at the very least make sure no admin account can log in that way.
3. Restrict network access
A PaaS, a vulnerability scanner or a video admin console has no business being published on the internet. Use a VPN, an IP allowlist or a jump host.
4. If the instance was exposed
Rotate the secrets the tool holds (API tokens, registry and cluster credentials for Kubero), and treat NextScan results as known to a third party.
It's the same pattern as the hard-coded secrets in LangChat, DataBuff, Simple Admin and Misaka Danmu last month: a default value shipped in the code ends up as a key shared with the whole world.
Why continuous monitoring matters for self-hosted tools
These tools often get installed outside the official inventory: a test PaaS, a scanner the security team spun up, an internal video platform. Nobody tracks their CVEs, and none of the three has an NVD CPE configuration yet, so a standard product search won't surface them.
With cveo.tech, add those tools to your inventory with their version and get an automatic alert as soon as a CVE that affects them is published, even before NIST analysis.