/

CVE-2026-55526

Severity:

HIGH RISK

(8.5)

Server-Side Request Forgery (CWE-918) in PraisonAI Spider

CVE-2026-55526

Deferred

Vulnerability type

Category:

Server-Side Request Forgery

CWE:

CWE-350, CWE-918

Impact:

Unauthorized Internal Network Access

Severity:

HIGH RISK

(8.5)

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

A server-side request forgery (SSRF) vulnerability in PraisonAI's spider tools allows specially crafted hostnames to bypass host validation and resolve to loopback or other internal addresses.

When an attacker can influence a URL processed by scrape_page, the service may make HTTP requests to internal services that should not be externally reachable.

Affected context

The issue affects PraisonAI deployments using the vulnerable spider_tools hostname validation and scrape_page fetching behavior. It becomes risky when user-controlled or otherwise untrusted URLs are accepted, the hostname is validated without first resolving it, and the resulting request can reach loopback, private, or otherwise restricted network destinations.

Exploit preconditions

  • Attack surface: The scrape_page or equivalent web-fetching functionality
  • Required access: The ability to provide or influence a URL submitted to the spider tool
  • Trigger point: Host validation followed by HTTP fetching
  • Why exploitation is possible: A hostname can appear to be external during validation while DNS resolution maps it to a loopback or internal address
  • Environmental dependency: The worker or application process must have network access to the resolved destination

Why this happens

The vulnerability results from validating a hostname as text without resolving ordinary hostnames before making the request. Hostnames that encode or resolve to loopback and private addresses can therefore pass an allow-or-deny check based on the unresolved name.

This maps primarily to CWE-918: Server-Side Request Forgery, because attacker-influenced URLs cause the server to initiate requests to unintended destinations. It also relates to CWE-350: Reliance on Reverse DNS Resolution, through unsafe assumptions about hostname and address identity during network validation.

Potential impact

  • An attacker may access internal HTTP services
  • Sensitive metadata or service responses could be disclosed
  • Internal administrative endpoints could be reached
  • The application may be used to scan or map reachable network resources
  • Downstream actions may be triggered against trusted internal services

General mitigation guidance

Mitigation generally focuses on validating the resolved destination rather than trusting the original hostname.

Common defensive patterns include:

  • Resolve hostnames before connection and reject loopback, link-local, private, multicast, and other restricted address ranges
  • Fail closed when DNS resolution or address classification cannot be completed
  • Revalidate the destination at connection time where DNS changes could create a mismatch
  • Prefer explicit outbound host and network allowlists over broad URL acceptance
  • Restrict the runtime's egress using network segmentation and least-privilege policies
  • Limit redirects and apply the same validation rules to every redirect target
  • Avoid exposing cloud metadata and internal control-plane endpoints to the fetching process

How teams detect issues like this

Organizations can detect SSRF risks through:

  • Static analysis rules for server-side HTTP clients that accept user-controlled URLs
  • Code review checks requiring DNS resolution and IP-range classification before requests
  • Tests using loopback, private, link-local, encoded, and DNS-mapped hostnames
  • Egress firewall and proxy policies that block restricted network ranges
  • Runtime logging of requested URLs, resolved addresses, redirects, and denied destinations
  • Automated security review tools that flag URL fetchers lacking destination validation

Sources

SourceLink
github.comgithub.com/MervinPraison/PraisonAI/commit/2f9677abb2ea68eab864ee8b6a828fd0141612e1
github.comgithub.com/MervinPraison/PraisonAI/releases/tag/v4.6.58
github.comgithub.com/MervinPraison/PraisonAI/security/advisories/GHSA-x44h-65qv-cw74

Frequently asked

questions

It may be remotely exploitable when an attacker can submit or influence URLs processed by the spider tool and the service can reach the resulting internal destination.

Not necessarily. The vulnerable application performs the request on the attacker's behalf, so the attacker generally needs access to the URL-fetching functionality rather than direct internal network access.

Resolve hostnames before making requests, reject resolved addresses in restricted ranges, fail closed on DNS errors, and apply equivalent checks to redirects and connection-time destinations.

Use strict outbound allowlists, enforce egress network controls, treat all fetched URLs as untrusted input, and test hostname validation against alternate DNS representations and private-address targets.

Find out what's already
exploitable in your codebase.