content-security-policy

Content-Security-Policy

Design and deploy CSP to prevent XSS and unauthorized resource loading.

Context

You are a senior security engineer implementing CSP for $ARGUMENTS. CSP is the browser's defense against XSS by controlling which scripts, styles, and resources can load. A well-designed CSP blocks inline scripts and restricts external resources to trusted domains, making XSS significantly harder to exploit.

Domain Context

• CSP Directives: default-src, script-src, style-src, img-src, font-src, connect-src, frame-src, form-action, base-uri
• Values: 'self' (same origin), 'none' (block all), domain whitelist, nonces (random tokens per request), hashes (SHA-256)
• Violation Reporting: report-uri or report-to for CSP violations; use to detect XSS attempts and policy issues
• Browser Support: Widely supported in modern browsers; older browsers ignore CSP (graceful degradation)

Instructions

1. Start with Report-Only Mode:

   • Deploy CSP with Content-Security-Policy-Report-Only header first
   • Set report-uri to collect violations
   • Monitor for 1-2 weeks; fix legitimate violations (misconfigured resources)
   • Then enforce with Content-Security-Policy header (report-only + enforcement can coexist)

2. Design Base Policy:

   • default-src 'self' — allow same-origin resources by default
   • script-src 'self' — allow same-origin scripts only (no inline scripts)
   • style-src 'self' — allow same-origin styles only (no inline styles)
   • img-src 'self' data: https: — allow same-origin, data URIs, and HTTPS images
   • font-src 'self' — allow same-origin fonts only
   • connect-src 'self' — allow same-origin API calls only

3. Allow Trusted External Resources:

   • If you need external CDNs: script-src 'self' cdn.jsdelivr.net
   • If you need Google Fonts: font-src 'self' fonts.googleapis.com; style-src 'self' fonts.googleapis.com
   • Always prefer HTTPS for external resources
   • Whitelist specific subdomains, not entire domains (e.g., trusted-cdn.com not *.example.com)

4. Secure Inline Scripts & Styles:

   • Option 1 (Nonces): Generate random nonce per request; include in <script nonce="abc123"> and CSP header
     • script-src 'nonce-abc123' — only that script with matching nonce executes
   • Option 2 (Hashes): Hash inline content; include in CSP header
     • script-src 'sha256-hash' — only scripts matching hash execute
   • Option 3 (Refactor): Move inline scripts to external files (best practice)
   • Never use 'unsafe-inline' or 'unsafe-eval' unless absolutely necessary (defeats XSS protection)

5. Monitor & Iterate:

   • Review CSP violations in logs; identify patterns (third-party analytics, ads, etc.)
   • Add trusted domains as needed, but conservatively
   • Tighten policy over time (remove unnecessary directives)
   • Test CSP in different browsers (Chrome, Firefox, Safari, Edge)
   • Document policy rationale for future maintainers

Anti-Patterns

• Setting default-src * or overly permissive policies; this provides no protection against XSS
• Using 'unsafe-inline' to fix CSP violations; refactor code instead; inline scripts defeat CSP's purpose
• Not using nonces/hashes; without them, you must rely on domain whitelisting, which is fragile
• Ignoring CSP violation reports; violations indicate XSS attempts or misconfigured resources
• Setting CSP on sensitive endpoints only; apply consistently across entire application

Further Reading

• Content-Security-Policy Spec: https://w3c.github.io/webappsec-csp/
• OWASP CSP Cheat Sheet: https://cheatsheetseries.owasp.org/cheatsheets/Content_Security_Policy_Cheat_Sheet.html
• MDN CSP Guide: https://developer.mozilla.org/en-US/docs/Web/HTTP/CSP
• CSP Violations Handling: https://developer.mozilla.org/en-US/docs/Web/API/SecurityPolicyViolationEvent