API7:2023 — Server Side Request Forgery

Slide 9 · Two Attack Patterns

Two patterns. Two different types of damage.

Cloud credential theft and internal network pivot — both enabled by one unvalidated URL field.

Pattern 1: Cloud Credential Theft

Target: cloud metadata endpoint (169.254.169.254)

What it returns: temporary IAM/GCP/Azure credentials

Effect: attacker acts as the application’s cloud identity

Worst case: all data the app role can access — S3 buckets, databases, secrets manager

Detection: no network anomaly (same server, same endpoint, GET request)

Pattern 2: Internal Network Pivot

Target: internal IPs (10.x.x.x, 192.168.x.x, internal hostnames)

What it returns: responses from services that trust the app server

Effect: attacker reaches services that assume all callers are trusted

Worst case: admin panels, Redis, Elasticsearch, internal APIs with no auth

Detection: unusual outbound connections in network logs (if monitored)

Both patterns exploit the same root cause

The server is trusted by cloud metadata endpoints and internal services because it’s on the network. That trust is designed for the application to use — not for attacker-controlled inputs to exploit. SSRF weaponizes legitimate network trust by making the server fetch attacker-chosen URLs.

← Back Pattern 1 mechanics →