On this page (7 sections)

A critical flaw tracked as CVE-2026-130840 affects BISHENG versions before 3.0.0. According to the GitHub Advisory Database entry, the application uses a hard-coded default jwt_secret to sign login tokens, so a remote attacker with no account can forge an administrator session and take over the instance.

If you run a self-hosted BISHENG deployment, especially one reachable from the internet or a shared network, treat this as an urgent upgrade and credential-rotation task. The advisory was published on November 17, 2026, and at the time of writing only one independent publisher covers it.

What happened

BISHENG signs its authentication tokens with a secret value called jwt_secret, handled by a component named AuthJwt. The advisory says the application relies on a hard-coded default value for that secret. Anyone who knows the default can produce tokens the server will accept as genuine.

The advisory describes the result plainly: an unauthenticated remote attacker can create an HS256 token offline that names user_id 1 with the admin role, present it as the access_token_cookie, and obtain full administrative access to users, workspaces, knowledge bases and workflows. No password, no user interaction and no prior foothold is needed. The CVSS v4 metrics on the advisory page reflect that: network attack vector, low complexity, no privileges required, and high impact on confidentiality, integrity and availability of the vulnerable system.

Why a hard-coded secret is so dangerous

A JSON Web Token (JWT) is a signed statement such as 'this user is an admin'. The server trusts it only because the signature can be produced solely by someone holding the secret. With HS256, the same secret is used to both sign and verify. If that secret is a known default, the signature proves nothing: every attacker already holds the signing key.

This is the weakness class CWE-321. It is not a bug in JWT itself, and it does not require breaking any cryptography. The failure is in key management: a value that must be unique and private was shipped as a constant in the source code. The project's settings file at tag v2.6.0-fix2 is one of the references listed by the advisory, and a fix commit is also referenced. We have not been able to confirm from the research exactly how the fix changes the behavior, so check the release notes for 3.0.0 before assuming what it does with existing configurations.

Who is affected

  • Any deployment of BISHENG older than 3.0.0 that still relies on the shipped default jwt_secret.
  • Instances reachable over the network by untrusted users, including internet-facing servers and flat internal networks.
  • Teams that store sensitive material in BISHENG knowledge bases or workflows, since an attacker with admin access can reach them.
The research does not say whether instances that had a custom jwt_secret configured are safe. The advisory describes the problem as abuse of the hard-coded default, which suggests a unique secret would change the picture, but this is not confirmed. Do not rely on it as a substitute for upgrading.

How to check your exposure

Start by finding out which version you run and whether your configuration overrides the default secret. Look at the application's release information or your deployment manifest (container image tag, git tag or package version) and compare it with 3.0.0.

Next, review how jwt_secret is set in your deployment. If it is absent from your configuration and environment, the application is probably using the built-in default. If you set it yourself, confirm it is long, random and not reused elsewhere. How BISHENG reads this setting is not described in the research, so confirm the exact mechanism in the BISHENG documentation.

Then look for signs of misuse. This read-only search runs on a Linux host and counts lines in your reverse proxy access log that mention the access_token_cookie name, which helps you find log lines to inspect. Replace the path with your own log location; many setups do not log cookies at all, in which case this will return 0 and is not proof of safety.

Count log lines mentioning the session cookie name (run on the proxy host; adjust the log path)Shell
grep -c 'access_token_cookie' /path/to/your/access.log

A more useful check is within the application itself: review the list of administrator accounts, recent changes to users, workspaces, knowledge bases and workflows, and any admin logins at times you cannot explain. Because the forged token names user_id 1, activity attributed to the first account deserves particular attention.

What to do now

  1. Identify every BISHENG instance you run and its version; anything before 3.0.0 is in scope.
  2. Upgrade to 3.0.0 or later after reading the release notes, since the advisory itself lists patched versions as unknown.
  3. Set a new, long, random jwt_secret unique to each deployment, stored in your secrets manager and not in the code repository, following the BISHENG documentation for how the value is configured.
  4. After changing the secret, restart the application if required and verify against the BISHENG documentation and release notes that tokens signed with the old key no longer validate. This should be the case if the application reads the new value, but the research does not confirm it.
  5. Restrict network access to the BISHENG admin interface using a VPN, IP allow list or authenticated reverse proxy until you are patched.
  6. Review admin accounts, users, workspaces, knowledge bases and workflows for changes you did not make.
  7. Rotate any API keys, model-provider credentials or database passwords that BISHENG workflows or knowledge bases could read.
  8. If you find signs of compromise, follow a structured response such as our guide on what to do when a website is hacked.
Caution. Changing the signing secret and restarting the service may log out users and interrupt running workflows. Plan it for a maintenance window, and take a backup of configuration and data first.
Tip. Generate the new secret with your operating system's cryptographically secure random source and give each environment (test, staging, production) its own value. A shared secret across environments recreates the same weakness.

Detection and monitoring afterwards

Forged tokens look like valid requests, so network signatures are unreliable. Rely on application audit trails: alert on new administrator accounts, role changes, exports of knowledge bases and edits to workflows. Keep reverse proxy logs for several weeks so you can investigate retroactively.

Exposure matters as much as the patch. If the admin interface does not need to be public, take it off the internet. A web application firewall can add a layer in front of an application like this, though it is not a replacement for upgrading; see what a WAF is and is not. For externally visible changes such as unexpected certificate or header changes on the host, Webforta Monitor can help you notice drift.

What we do not know yet

The sources available are the GitHub Advisory Database entry, which is marked unreviewed, and the NVD record, which repeats the same text. The advisory also references a VulnCheck write-up. We have no confirmation of in-the-wild exploitation, no named victims and no official vendor statement in the research. Treat the absence of exploitation reports as unknown rather than reassuring, because the attack needs no special skill once the default secret is known.

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-2026-130840 - GitHub Advisory Databasegithub.com
  2. CVE-2026-130840: BISHENG before 3.0.0 contains an authentication bypass vulnerability that allows unauthenticated remote attackers to impersonate administrators by abusing a…nvd.nist.gov ·
  • CVE-2026-130840
  • BISHENG
  • JWT
  • authentication bypass
  • hard-coded credentials