latestaiagents/injection-prevention

Injection Prevention (OWASP A01)

Prevent SQL, NoSQL, Command, and other injection attacks by validating and sanitizing all user input.

When to Use

• Reviewing code that builds SQL/NoSQL queries
• Code that executes shell commands
• Any place user input reaches an interpreter
• Building APIs that accept user data
• Migrating from string concatenation to parameterized queries

Injection Types

| Type | Danger | Common Locations | |------|--------|------------------| | SQL Injection | CRITICAL | Database queries, ORMs with raw queries | | NoSQL Injection | CRITICAL | MongoDB, Redis, Elasticsearch queries | | Command Injection | CRITICAL | Shell exec, system calls, child_process | | LDAP Injection | HIGH | Directory service queries | | XPath Injection | HIGH | XML document queries | | Expression Language | HIGH | Template engines, eval() |

Detection Patterns

SQL Injection Red Flags

// VULNERABLE - String concatenation
const query = "SELECT * FROM users WHERE id = " + userId;
const query = `SELECT * FROM users WHERE name = '${userName}'`;

// VULNERABLE - Format strings
query = "SELECT * FROM users WHERE id = %s" % user_id

// VULNERABLE - String interpolation in ORM
User.where("name = '#{params[:name]}'")

Command Injection Red Flags

// VULNERABLE - Direct user input in commands
exec(`ls ${userInput}`);
system("ping " + ipAddress);
child_process.exec(`convert ${filename} output.png`);

// VULNERABLE - eval with user data
eval(userCode);
new Function(userInput)();

Prevention Techniques

1. Parameterized Queries (SQL)

// SAFE - Node.js with parameterized query
const result = await db.query(
  'SELECT * FROM users WHERE id = $1 AND status = $2',
  [userId, status]
);

// SAFE - Using ORM properly
const user = await User.findOne({ where: { id: userId } });

// SAFE - Prepared statements
const stmt = db.prepare('SELECT * FROM users WHERE email = ?');
const user = stmt.get(email);

# SAFE - Python with parameterized query
cursor.execute(
    "SELECT * FROM users WHERE id = %s AND status = %s",
    (user_id, status)
)

# SAFE - SQLAlchemy ORM
user = session.query(User).filter(User.id == user_id).first()

// SAFE - PHP PDO with prepared statements
$stmt = $pdo->prepare('SELECT * FROM users WHERE id = :id');
$stmt->execute(['id' => $userId]);

2. NoSQL Injection Prevention

// VULNERABLE - MongoDB with user object
db.users.find({ username: req.body.username, password: req.body.password });
// Attack: { "username": "admin", "password": { "$ne": "" } }

// SAFE - Validate and sanitize input
const username = String(req.body.username).slice(0, 50);
const password = String(req.body.password);

// SAFE - Use mongoose with schema validation
const userSchema = new Schema({
  username: { type: String, required: true, maxlength: 50 },
  password: { type: String, required: true }
});

3. Command Injection Prevention

// VULNERABLE
exec(`convert ${filename} output.png`);

// SAFE - Use array arguments (no shell)
execFile('convert', [filename, 'output.png']);

// SAFE - Whitelist allowed values
const ALLOWED_FORMATS = ['png', 'jpg', 'gif'];
if (!ALLOWED_FORMATS.includes(format)) {
  throw new Error('Invalid format');
}

// SAFE - Use library instead of shell
const sharp = require('sharp');
await sharp(inputFile).toFile(outputFile);

# VULNERABLE
os.system(f"ping {ip_address}")

# SAFE - Use subprocess with list arguments
import subprocess
subprocess.run(['ping', '-c', '4', ip_address], check=True)

# SAFE - Use library instead of shell
import socket
socket.gethostbyname(hostname)

4. Input Validation

// Validation helper
function validateInput(input, options = {}) {
  const { maxLength = 255, pattern, allowedValues } = options;

  // Type check
  if (typeof input !== 'string') {
    throw new ValidationError('Input must be a string');
  }

  // Length check
  if (input.length > maxLength) {
    throw new ValidationError(`Input exceeds ${maxLength} characters`);
  }

  // Pattern check
  if (pattern && !pattern.test(input)) {
    throw new ValidationError('Input contains invalid characters');
  }

  // Whitelist check
  if (allowedValues && !allowedValues.includes(input)) {
    throw new ValidationError('Input value not allowed');
  }

  return input;
}

// Usage
const userId = validateInput(req.params.id, {
  maxLength: 36,
  pattern: /^[a-f0-9-]+$/i  // UUID pattern
});

Code Review Checklist

Before Approving Code, Verify:

• [ ] No string concatenation in SQL queries
• [ ] All database queries use parameterized statements
• [ ] User input is validated before use
• [ ] Shell commands use array arguments, not string interpolation
• [ ] No eval(), new Function(), or similar with user data
• [ ] Input length limits are enforced
• [ ] Type checking is performed on all inputs
• [ ] Error messages don't expose query structure

Testing for Injection

Manual Test Payloads

# SQL Injection tests
' OR '1'='1
'; DROP TABLE users; --
' UNION SELECT * FROM passwords --
1; SELECT * FROM users

# NoSQL Injection tests
{"$gt": ""}
{"$ne": null}
{"$where": "sleep(5000)"}

# Command Injection tests
; ls -la
| cat /etc/passwd
`whoami`
$(id)

Automated Tools

# SQLMap for SQL injection
sqlmap -u "http://target.com/page?id=1" --dbs

# NoSQLMap for NoSQL injection
nosqlmap -u "http://target.com/api/users"

# Commix for command injection
commix -u "http://target.com/ping?ip=127.0.0.1"

Framework-Specific Guidance

Express.js / Node.js

• Use express-validator for input validation
• Use Sequelize/Prisma ORM with proper escaping
• Never use eval() or child_process.exec() with user input

Django / Python

• Use Django ORM (automatic parameterization)
• Use subprocess.run() with list arguments
• Enable SECURE_CONTENT_TYPE_NOSNIFF

Rails / Ruby

• Use ActiveRecord properly (automatic parameterization)
• Avoid where("column = '#{value}'")
• Use Shellwords.escape() for shell arguments

Laravel / PHP

• Use Eloquent ORM or Query Builder
• Use PDO prepared statements
• Use escapeshellarg() for shell arguments

Best Practices

1. Defense in Depth: Combine input validation + parameterized queries + output encoding
2. Least Privilege: Database accounts should have minimum required permissions
3. Whitelist Over Blacklist: Define what's allowed, not what's forbidden
4. Escape Output: Even with safe queries, escape data when displaying
5. Log Suspicious Input: Monitor for injection attempts
6. Regular Security Audits: Use automated scanners in CI/CD pipeline