How Attackers Exploit SSRF (Server-Side Request Forgery)

How Attackers Exploit SSRF (Server-Side Request Forgery)

1 min read

SSRF · cloud security · Injection · OWASP

Authors

PandaONE Team

Why URL-fetch features become SSRF, how cloud metadata and internal hosts get reached, and what proof-backed testing looks for — authorize-only education.

Your PDF preview service fetches whatever URL the user pastes. Point it at http://169.254.169.254/ and the response body is not a PDF anymore.

SSRF tricks your server into making requests on an attacker’s behalf — often to cloud metadata, internal admin panels, or private network hosts. Features that “fetch a URL for the user” (previews, webhooks, importers, PDF renderers) are the usual door.

Educational only; test authorized targets. Full walkthrough: Learn → SSRF.

The abuse pattern

  1. Find a parameter that causes the backend to request a URL.
  2. Point it at an internal destination (http://169.254.169.254/, http://localhost:8080/admin, a staging hostname).
  3. Read the response (full SSRF) or infer success via timing/errors (blind SSRF).
  4. Chain with redirects, DNS rebinding, or alternate encodings to bypass naive denylists.

In cloud environments, metadata credentials can turn SSRF into broader account compromise.

Why denylists fail

Filtering the string 169.254.169.254 misses IPv6, decimal forms, DNS names that resolve privately, and redirect hops to blocked IPs after an allowlisted first hop. Prefer allowlists, hardened resolvers, and network egress policy.

What proof looks like

Under rules of engagement, proof documents that the application initiated a fetch to an unintended or attacker-controlled destination and returned evidence of that behavior. “URL parameter present” without a demonstrated fetch is scanner noise. PandaONE classifies confirmed SSRF under Injection.

Prevention checklist

  • Allowlist destinations for server-side HTTP
  • Block private/link-local/metadata ranges after your own DNS resolution
  • Disable dangerous redirects and schemes
  • Put fetches behind a locked-down egress proxy
  • Never echo raw arbitrary remote bodies to end users

Related: IDOR abuse, open redirect class, attack surface management. Exercise fetch features with AI pentest on a verified domain.

Featured on ScrollLaunchFeatured on CodeHype