/

CVE-2026-67620

Severity:

HIGH RISK

(7.7)

SSRF (CWE-918) in Flowise Cloud Metadata Protection

CVE-2026-67620

Analyzed

Vulnerability type

Category:

Server-Side Request Forgery

CWE:

CWE-918

Impact:

Cloud Metadata and Credential Exposure

Severity:

HIGH RISK

(7.7)

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 through version 3.1.4 contains a server-side request forgery flaw in its URL-fetching security guard, allowing requests to reach omitted Oracle Cloud Infrastructure and Alibaba Cloud metadata endpoints. An attacker with access to the fetch-links API, or potentially an unauthenticated user when a public chatflow exposes URL-fetching functionality, may obtain instance identity information and role credentials.

Affected context

The vulnerability occurs in Flowise deployments that expose URL-fetching functionality through the fetch-links API or public chatflows. Risk is greatest when the Flowise server runs on Oracle Cloud Infrastructure or Alibaba Cloud and can access instance metadata services, while the SSRF guard relies on an incomplete deny list and does not consistently prevent redirect-based access to restricted destinations.

Exploit preconditions

  • Attack surface: The Flowise fetch-links API or a public chatflow containing

URL-fetching functionality.

  • Required access: Normally authenticated access to the relevant API; access

may be unauthenticated when URL-fetching nodes are exposed through a public

chatflow.

  • Trigger point: Server-side retrieval of a caller-controlled URL. - Why exploitation is possible: The deny list omits cloud metadata addresses

used by Oracle Cloud Infrastructure and Alibaba Cloud, and redirect

handling can allow requests to bypass destination validation.

Why this happens

The issue is caused by an incomplete SSRF defense that treats a deny list as sufficient protection for server-side URL retrieval. Because sensitive metadata destinations are not comprehensively identified and redirects may be evaluated inadequately, attacker-controlled URLs can cause the server to access internal cloud services. This behavior maps to CWE-918, where an application retrieves a remote resource based on attacker-influenced input without reliably restricting the destination.

Potential impact

  • An attacker may retrieve cloud instance identity information.
  • Cloud role credentials or temporary access tokens could be exposed.
  • Compromised credentials could enable unauthorized access to cloud resources.
  • Internal services reachable from the Flowise host may be probed.
  • Affected deployments could experience data exposure, persistence, or lateral movement.

General mitigation guidance

Mitigation generally focuses on reducing trust in user-supplied URLs and enforcing destination controls throughout the complete request lifecycle. Secure-by-default patterns include using an allowlist of required external destinations, rejecting private, link-local, loopback, and cloud metadata address ranges, resolving and validating every redirect target, and rechecking destinations after DNS resolution. Network-level egress controls can block metadata services independently of application logic. URL-fetching features can be disabled when unnecessary, and Flowise services should run with minimal cloud permissions that exclude metadata-derived credentials where possible.

How teams detect issues like this

Detection approaches include static analysis for server-side HTTP requests built from request parameters, review of deny-list-only SSRF defenses, and checks for redirect handling that does not revalidate destinations. Code reviews should examine DNS resolution, IP-range filtering, URL scheme restrictions, and cloud metadata protections. Runtime monitoring can alert on requests from Flowise to link-local, private, or known metadata addresses, while network policy can enforce egress restrictions and log blocked attempts. Automated security review tools may help identify common SSRF patterns but should be supplemented with deployment-level testing.

Sources

SourceLink
flowiseai.comflowiseai.com/sunset
github.comgithub.com/abdugafforov-bobur/CVE-2026-67620-poc
vulncheck.comvulncheck.com/advisories/flowise-ssrf-via-fetch-links-endpoint-incomplete-deny-list

Frequently asked

questions

It can be remotely exploitable when an attacker can reach the Flowise URL-fetching functionality. Authentication may be required, although public chatflows with URL-fetching nodes can broaden exposure.

The attacker needs access to the fetch-links API or another exposed feature that causes the server to retrieve a supplied URL. The required access depends on the deployment's authentication and chatflow configuration.

Use strict destination allowlisting, block cloud metadata and private address ranges, validate resolved addresses, and revalidate every redirect. Independent network egress controls provide an additional protection layer.

Treat all URL targets as untrusted, avoid deny-list-only designs, restrict supported schemes and destinations, monitor outbound requests, and assign the application only the cloud permissions it requires.

Find out what's already
exploitable in your codebase.