Security Audit

Security Audit Skill

Systematically audit code for common security vulnerabilities.

Workflow

1. Identify Attack Surface

Map entry points:

• API endpoints
• User input forms
• File uploads
• URL parameters
• Headers and cookies
• WebSocket messages

# Find API routes
grep -r "app.get\|app.post\|router\." --include="*.ts" --include="*.js"

# Find form handlers
grep -r "onSubmit\|handleSubmit" --include="*.tsx" --include="*.jsx"

2. Check OWASP Top 10

Go through each category systematically:

3. Review Code Patterns

Search for dangerous patterns:

# SQL injection risks
grep -r "query.*\${" --include="*.ts"
grep -r "execute.*+" --include="*.py"

# Command injection risks
grep -r "exec\|spawn\|system" --include="*.ts" --include="*.js"

# XSS risks
grep -r "innerHTML\|dangerouslySetInnerHTML" --include="*.tsx"

# Hardcoded secrets
grep -r "password\|secret\|api_key\|token" --include="*.ts" --include="*.env*"

4. Check Dependencies

# JavaScript
npm audit
npx snyk test

# Python
pip-audit
safety check

5. Document Findings

For each vulnerability:

• Severity (Critical/High/Medium/Low)
• Location (file:line)
• Description
• Proof of concept (if safe)
• Recommended fix

Quick Reference

Common Vulnerabilities

Injection (SQL, NoSQL, Command, LDAP)

Vulnerable:

// SQL injection
const query = `SELECT * FROM users WHERE id = ${userId}`;

// Command injection
exec(`convert ${userFilename} output.png`);

// NoSQL injection
User.find({ username: req.body.username });

Fixed:

// Parameterized query
const query = 'SELECT * FROM users WHERE id = ?';
db.query(query, [userId]);

// Validate/escape input
const safeFilename = path.basename(userFilename);
exec(`convert ${shellescape([safeFilename])} output.png`);

// Sanitize NoSQL input
User.find({ username: String(req.body.username) });

Cross-Site Scripting (XSS)

Vulnerable:

// Reflected XSS
element.innerHTML = userInput;

// Stored XSS
<div dangerouslySetInnerHTML={{ __html: comment }} />

Fixed:

// Use textContent
element.textContent = userInput;

// Sanitize HTML
import DOMPurify from 'dompurify';
<div dangerouslySetInnerHTML={{ __html: DOMPurify.sanitize(comment) }} />

Broken Authentication

Vulnerable:

// Weak session
const sessionId = Math.random().toString();

// No rate limiting on login
app.post('/login', (req, res) => { /* no limit */ });

// Password in URL
window.location = `/login?password=${password}`;

Fixed:

// Strong session ID
const sessionId = crypto.randomBytes(32).toString('hex');

// Rate limiting
app.post('/login', rateLimit({ max: 5, windowMs: 15*60*1000 }), ...);

// POST body for credentials
fetch('/login', { method: 'POST', body: JSON.stringify({ password }) });

Sensitive Data Exposure

Vulnerable:

// Logging sensitive data
console.log('User login:', { email, password });

// Error details to client
res.status(500).json({ error: err.stack });

// Hardcoded secrets
const API_KEY = 'sk-1234567890abcdef';

Fixed:

// Sanitize logs
console.log('User login:', { email, password: '[REDACTED]' });

// Generic error
res.status(500).json({ error: 'Internal server error' });

// Environment variables
const API_KEY = process.env.API_KEY;

Security Headers Checklist

// Required headers
helmet({
  contentSecurityPolicy: true,      // XSS protection
  hsts: true,                       // Force HTTPS
  noSniff: true,                    // MIME type sniffing
  frameguard: { action: 'deny' },   // Clickjacking
  xssFilter: true                   // XSS filter
});

Authentication Checklist

• Passwords hashed with bcrypt/argon2
• Session tokens are cryptographically random
• Sessions expire and can be invalidated
• MFA available for sensitive operations
• Password requirements enforced
• Account lockout after failed attempts
• Secure cookie flags (HttpOnly, Secure, SameSite)

Authorization Checklist

• Every endpoint checks permissions
• IDs in URLs are validated against user
• Elevation of privilege paths reviewed
• Admin functions separated
• Default deny policy

See Also

• owasp.md - Full OWASP Top 10 checklist
• scripts/check_secrets.py - Secret scanning script