NewApp Monitoring is liveSign up free
AgentbeamBeta
AI agent vulnerability · Langflow

CVE-2026-33053: Langflow is Missing Ownership Verification in API Key Deletion (IDOR)

highCVSS 4.0 7.1GHSA-rf6x-r45m-xv3w

Langflow's DELETE /api/v1/api_key/{api_key_id} endpoint deleted any API key by ID without checking who owned it. Any logged-in user could delete another user's keys. The advisory data lists langflow before 1.9.0 as vulnerable and 1.9.0 as patched. The fix commit makes the delete function compare the key's owner to the caller.

Published Mar 18, 2026 · Updated Jun 6, 2026 · Analysis Oct 31, 2026

What an attacker can do

The route in src/backend/base/langflow/api/v1/api_key.py only required an authenticated user through the get_current_active_user dependency. It never passed that user to delete_api_key() in crud.py. That function loaded the key by UUID and deleted it. This is an insecure direct object reference (CWE-639). The GET and POST routes in the same file already filter by user_id, so only delete was missing the check.

An attacker needs a valid Langflow account and the UUID of a victim's key. The advisory says the IDs can be guessed or discovered. Deleting keys breaks other users' integrations: any application or script that calls the Langflow API with that key fails at once. The advisory calls this account takeover and denial of service, but the flaw only deletes keys. It does not reveal or create them. The CVSS 4.0 vector reflects that, with availability high and no confidentiality or integrity impact.

Teams that run a shared Langflow server for several users, or expose flows to production apps through API keys, are the ones to worry about.

Who is affected

The pip package langflow, versions before 1.9.0 per the advisory data. A release tag 1.7.2 is also linked as a reference, so treat 1.9.0 as the safe target. Check the installed version:

pip show langflow

Multi-user deployments, where several accounts share one server, are the ones exposed in practice.

langflow

Ecosystem
PyPI
Affected
< 1.9.0
Fixed in
1.9.0
Check
pip show langflow

How to fix it

Upgrade to 1.9.0:

pip install --upgrade "langflow>=1.9.0"

The fix commit changes the route to take the current user and pass current_user.id to delete_api_key(). That function now raises "API Key not found" when the owner does not match, the same message as for a missing key, so the response does not reveal that another user's key exists. The sources give no workaround. If you cannot upgrade, limit who can log in to the instance.

How to tell if you were hit

Review Langflow logs for DELETE requests to /api/v1/api_key/ followed by an id, and compare the requesting user with the key's owner. If integrations suddenly start failing with authentication errors, check whether their keys still exist in the Langflow API keys page. Recreate any missing key and rotate the others.

Original advisory

From the GitHub Advisory Database (CC BY 4.0).

Detection Method: Kolega.dev Deep Code Scan

| Attribute | Value |
| Location | src/backend/base/langflow/api/v1/api_key.py:44-53 |
| Practical Exploitability | High |
| Developer Approver | [email protected] |

Description

The delete_api_key_route() endpoint accepts an api_key_id path parameter and deletes it with only a generic authentication check (get_current_active_user dependency). However, the delete_api_key() CRUD function does NOT verify that the API key belongs to the current user before deletion.

Affected Code

@router.delete("/{api_key_id}", dependencies=[Depends(auth_utils.get_current_active_user)])
async def delete_api_key_route(
    api_key_id: UUID,
    db: DbSession,
):
    try:
        await delete_api_key(db, api_key_id)
    except Exception as e:
        raise HTTPException(status_code=400, detail=str(e)) from e
    return {"detail": "API Key deleted"}

Evidence

In crud.py lines 44-49, delete_api_key() retrieves the API key by ID and deletes it without checking if the key belongs to the authenticated user. The endpoint also doesn't pass the current_user to the delete function for verification.

Impact

An authenticated attacker can enumerate and delete API keys belonging to other users by guessing or discovering their API key IDs. This allows account takeover, denial of service, and disruption of other users' integrations.

Recommendation

Modify the delete_api_key endpoint and function: (1) Pass current_user to the delete function; (2) In delete_api_key(), verify api_key.user_id == current_user.id before deletion; (3) Raise a 403 Forbidden error if the user doesn't own the key. Example: if api_key.user_id != user_id: raise HTTPException(status_code=403, detail='Unauthorized')

Notes

Confirmed IDOR vulnerability. The delete_api_key_route endpoint at line 44-53 accepts an api_key_id and calls delete_api_key(db, api_key_id) without passing the current_user. The CRUD function delete_api_key() at crud.py:44-49 retrieves the API key by ID and deletes it without verifying ownership (api_key.user_id == current_user.id). Compare this to the GET endpoint at lines 17-28 which correctly filters by user_id, and the POST endpoint at lines 31-41 which correctly associates the key with user_id. An authenticated attacker can delete any user's API keys by guessing/enumerating UUIDs. Fix: Pass current_user to delete_api_key and verify api_key.user_id == current_user.id before deletion, returning 403 if unauthorized.

Developer Review Notes

Does not accept current_user as a parameter. Allowing deletion of any user's API keys even without permissions.

Credited to FaizanKolega, kolega-ai-dev, andifilhohub, erichare.

Questions

Does CVE-2026-33053 need an account?

Yes. The attacker must be an authenticated Langflow user, and needs the UUID of the target key. The advisory says IDs could be guessed or discovered.

Can an attacker read someone else's API key with this flaw?

No. The affected endpoint deletes keys only. The impact is lost keys and broken integrations, not disclosure.

Which Langflow version fixes CVE-2026-33053?

1.9.0, per the advisory data. Versions before 1.9.0 are listed as vulnerable.

References

How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. The analysis is written by an AI model from the sources listed above and checked by automated rules before it is published. Check versions against the vendor's advisory before you act. Methodology

More Langflow vulnerabilities

All Langflow entries