/

CVE-2026-90535

Severity:

HIGH RISK

(7.5)

Missing Authorization (CWE-862) in Flowise TTS Endpoint

CVE-2026-90535

Analyzed

Vulnerability type

Category:

Missing Authorization

CWE:

CWE-862

Impact:

Targeted Denial of Service

Severity:

HIGH RISK

(7.5)

CVSS version:

3.1

Catch this CVE while code is being written

Detect vulnerable code directly in pull requests, so CVEs are fixed before merge.

Summary

Flowise versions before 3.1.4 contain a missing authorization check in the text-to-speech abort endpoint. An unauthenticated attacker who knows valid chatflow and chat identifiers may terminate another user's active prediction, causing targeted denial of service.

Affected context

The vulnerability occurs in Flowise deployments that expose the /api/v1/text-to-speech/abort endpoint and use user-supplied chatflowId and chatId values to identify active predictions. The risk exists when the endpoint does not verify that the requester owns or is otherwise authorized to control the referenced chatflow and chat session.

Exploit preconditions

  • Attack surface: The HTTP text-to-speech abort API endpoint
  • Required access: No authentication may be required
  • Required knowledge: A valid or discoverable chatflow identifier and chat identifier
  • Trigger point: A request to abort an active text-to-speech or chatflow prediction
  • Why exploitation is possible: The endpoint accepts object identifiers without enforcing ownership or equivalent authorization checks

Why this happens

The underlying bug is a failure to enforce authorization at an operation that changes the state of another user's active process. Identifiers supplied by the requester are treated as sufficient authority to cancel a prediction, rather than being checked against the authenticated principal or an appropriate access-control policy. This behavior maps to CWE-862: Missing Authorization, in which a protected operation lacks the permission validation needed to prevent unauthorized actions.

Potential impact

  • An attacker may terminate active predictions belonging to other users
  • Customer workflows could be interrupted
  • Applications could experience targeted service disruption
  • Repeated requests could reduce service availability or reliability
  • Operational teams may face increased support and recovery workload

General mitigation guidance

Mitigation generally focuses on enforcing authorization for every state-changing endpoint and reducing trust in caller-supplied identifiers.

Common defensive patterns include:

  • Require authentication where the operation affects user-owned or tenant-owned resources
  • Verify ownership, tenant scope, or explicit permission before aborting a prediction
  • Derive resource relationships from trusted server-side context rather than accepting them independently from the client
  • Apply deny-by-default authorization policies to API actions
  • Use least-privilege service identities and isolate tenant resources
  • Rate-limit repeated abort requests and expose only the minimum necessary endpoint behavior

How teams detect issues like this

Organizations commonly detect missing authorization through:

  • Static analysis rules for state-changing endpoints that consume object identifiers without permission checks
  • Code review of controller and route handlers for authentication, ownership, and tenant-boundary validation
  • Policy checks requiring authorization middleware on mutation and cancellation operations
  • Integration tests that attempt cross-user access to the same resource identifiers
  • API monitoring for unauthenticated abort requests, repeated cancellations, and unusual identifier patterns

Automated security review tools can help identify endpoints where resource access is not consistently tied to the requesting principal.

Sources

SourceLink
github.comgithub.com/FlowiseAI/Flowise/security/advisories/GHSA-xhxx-56g3-mx2r
vulncheck.comvulncheck.com/advisories/flowise-before-3.1.4-denial-of-service-via-text-to-speech-abort

Frequently asked

questions

Yes. An attacker may exploit the exposed API remotely when the endpoint is reachable and the required chatflow and chat identifiers are known or discoverable.

The described issue is unauthenticated, so a valid account may not be required. Deployment-specific access controls, network restrictions, or API gateways can affect practical exposure.

Enforce authentication and verify that the requester owns or is authorized to control the referenced chatflow and chat session before performing the abort operation.

Apply deny-by-default authorization consistently to state-changing APIs, bind resource checks to trusted server-side identity and tenant context, and test cross-user access during code review and integration testing.

Find out what's already
exploitable in your codebase.