CVE-2026-70475: 
Flowise vulnerability analysis and mitigation

CVE-2026-70475 is a Missing Authorization vulnerability in FlowiseAI Flowise, a drag-and-drop UI for building customized LLM workflows. The PUT /api/v1/executions/:id endpoint in packages/server/src/routes/executions/index.ts lacks the checkAnyPermission() middleware present on all other execution endpoints, allowing any authenticated user to modify execution records regardless of their assigned role. Affected versions are all releases up to and including 3.1.2 (npm package flowise); version 3.1.3 contains the fix. The vulnerability was discovered by researcher Dimpal Jadhav on 2026-03-17, published to the GitHub Advisory Database on 2026-08-04, and carries a CVSS v4.0 base score of 7.1 (High) (Github Advisory, Feedly).

The root cause is CWE-862 (Missing Authorization): the route definition for PUT /api/v1/executions/:id in packages/server/src/routes/executions/index.ts omits the checkAnyPermission() middleware call that guards the GET (executions:view) and DELETE (executions:delete) routes on the same router. An attacker with any valid authenticated session or API key can send a crafted HTTP PUT request directly to this endpoint, supplying arbitrary JSON to overwrite execution state, data, and metadata fields. No special configuration or elevated privileges beyond basic authentication are required, and no user interaction is needed (Github Advisory, Fix PR).

Vulnerable code (before fix):

// Line 11 - PUT has NO permission check
router.put(['/', '/:id'], executionController.updateExecution) // <-- MISSING checkAnyPermission

Fixed code:

router.put(['/', '/:id'], checkAnyPermission('executions:update'), executionController.updateExecution)

Successful exploitation allows a low-privileged authenticated user to perform unauthorized modification of any execution record within their workspace, constituting a privilege escalation. An attacker can overwrite execution state (e.g., marking a running workflow as FINISHED), inject or tamper with execution data and metadata, and manipulate workflow results — undermining the integrity of LLM-driven automation pipelines. Confidentiality and availability of the vulnerable system are not directly impacted, but data integrity of workflow executions is fully compromised for any execution ID the attacker can reference (Github Advisory).

A proof-of-concept reproduction command is publicly documented in the GitHub Security Advisory, making exploitation straightforward for any authenticated user (Github Advisory). The NVD SSVC assessment classifies exploitation status as poc (PoC available). The EPSS score is approximately 0.296% (22nd percentile), indicating a relatively low but non-negligible probability of exploitation in the wild within 30 days. No in-the-wild exploitation, threat actor attribution, or CISA KEV catalog listing has been reported at this time (Github Advisory).

1. Obtain valid credentials: Acquire any authenticated session token or low-privileged API key for the target Flowise instance (e.g., via a free/trial account or compromised low-privilege user).
2. Identify a target execution ID: Use the GET /api/v1/executions endpoint (which requires executions:view permission) or enumerate execution IDs through other means available to the authenticated user.
3. Craft a malicious PUT request: Construct an HTTP PUT request targeting PUT /api/v1/executions/<EXECUTION_ID> with a JSON body containing the desired manipulated fields:

curl -X PUT https://TARGET/api/v1/executions/EXECUTION_ID \
  -H "Authorization: Bearer LOW_PRIV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"state": "FINISHED", "data": "MANIPULATED"}'

4. Achieve unauthorized modification: Because the endpoint lacks checkAnyPermission() middleware, the server processes the request without verifying the caller's permissions, overwriting the execution record with attacker-controlled values.
5. Leverage tampered results: Use the manipulated execution state or data to influence downstream workflow logic, deceive administrators reviewing execution history, or inject malicious data into LLM pipeline outputs (Github Advisory).

• Network: Unexpected HTTP PUT requests to /api/v1/executions/<id> originating from user accounts that do not normally perform update operations; requests from low-privileged API keys to this endpoint.
• Logs: Server access logs showing PUT /api/v1/executions/ requests from accounts without executions:update role assignments (on patched systems, these would be rejected; on unpatched systems, they succeed silently); unusual frequency of PUT requests to the executions endpoint.
• Application Data: Execution records with unexpected state transitions (e.g., RUNNING → FINISHED without corresponding workflow completion events), or data/metadata fields containing values inconsistent with the original workflow run.

Upgrade the Flowise npm package to version 3.1.3 or later, which adds the checkAnyPermission('executions:update') middleware to the PUT /api/v1/executions/:id route and introduces a new executions:update RBAC permission (Flowise Release, Fix Commit). No official configuration-based workaround is available for unpatched versions; as an interim measure, restrict network access to the Flowise API to trusted users only and audit API key assignments to minimize the number of authenticated users who could exploit this endpoint. Review execution records for unexpected modifications as part of incident response.