Summary
Flowise document loader nodes using Cheerio, Playwright, or Puppeteer can be abused to send server-side requests to attacker-selected URLs when SSRF protections are bypassed.
An attacker who can submit or control a document URL may access cloud metadata services, internal applications, and private network resources, with retrieved responses exposed as document text.
Affected context
The vulnerability occurs in Flowise deployments that expose document loader functionality using Cheerio, Playwright, or Puppeteer and allow attacker-controlled URLs to be processed. Risk is higher when the application can reach cloud metadata endpoints, private network services, administrative interfaces, or other resources that are not publicly accessible.
Exploit preconditions
- Attack surface: Flowise document loader nodes that retrieve remote content
- Required access: Ability to create, modify, or trigger a workflow containing a document URL
- Trigger point: Server-side fetching and parsing of the supplied URL
- Why exploitation is possible: SSRF controls can be bypassed, allowing the server-side runtime to connect to internal, private, or cloud-provider endpoints and return response content to the workflow
Why this happens
The underlying bug is insufficient validation and enforcement of destination restrictions for server-side URL requests. When document loaders accept attacker-influenced URLs without reliably blocking private, loopback, link-local, or otherwise restricted destinations, the application becomes an unintended proxy into networks accessible from the server. This behavior maps to CWE-918: Server-Side Request Forgery.
Potential impact
- Exposure of cloud instance metadata or temporary credentials
- Unauthorized access to internal services and private network resources
- Disclosure of administrative or application data
- Network reconnaissance from the Flowise runtime
- Potential follow-on compromise if retrieved secrets enable privileged access
General mitigation guidance
Mitigation generally focuses on reducing trust in user-supplied URLs and enforcing destination controls at multiple layers.
Common defensive patterns include:
- Permit-listing required schemes, hosts, and ports instead of relying only on deny lists
- Blocking loopback, link-local, private, reserved, and cloud metadata address ranges
- Resolving hostnames safely and validating the resulting addresses before connection
- Rechecking destinations after redirects and disabling unnecessary redirects
- Isolating document-fetching workers from sensitive network segments
- Applying outbound firewall and egress policies with least privilege
- Limiting response size, connection scope, and accessible protocols
- Avoiding exposure of fetched response content when it is not required by the workflow
How teams detect issues like this
Detection approaches can include:
- Static analysis for document-fetching code that accepts user-controlled URLs without centralized validation
- Code review checks for incomplete SSRF filters, DNS resolution gaps, redirect handling weaknesses, and parser-specific bypasses
- Runtime monitoring of outbound requests to private, loopback, link-local, and metadata address ranges
- Egress policy alerts for unexpected protocols, ports, destinations, or sudden access to internal services
- Automated security testing using SSRF test cases across Cheerio, Playwright, and Puppeteer execution paths
- Policy enforcement requiring approved URL-fetching components and destination allowlists

