Security Model
Executive Summary
RHEA implements defense-in-depth security through Content Security Policy (CSP) enforcement, HTML sanitization via a shared sanitizer and DOMPurify, dependency vulnerability scanning via Renovate, and strict inline script control via nonce-based CSP. The security model covers the CDN edge, application layer, content layer, and supply chain.
Security Layers
| Layer | Control | Description |
|---|---|---|
| CDN Edge | AEM Edge Delivery CDN | TLS termination, DDoS protection, edge caching, geographic distribution |
| Content Security Policy | head.html CSP header | Strict script and resource loading policy |
| HTML Sanitization | sanitize.js + DOMPurify | XSS prevention for authored and fetched content |
| Dependency Management | Renovate + npm audit | Automated dependency updates and vulnerability detection |
| Access Control | GitHub + AEM IAM | Repository access, authoring permissions |
Content Security Policy
RHEA enforces a strict CSP defined in head.html:
script-src 'nonce-aem' 'strict-dynamic' 'unsafe-inline' http: https:;
base-uri 'self';
object-src 'none';
CSP Directives
| Directive | Value | Purpose |
|---|---|---|
| script-src | 'nonce-aem' 'strict-dynamic' | Only scripts with the aem nonce or loaded by trusted scripts are executed |
| script-src | 'unsafe-inline' | Fallback for browsers that don't support nonce (ignored when nonce is present) |
| script-src | http: https: | Fallback for browsers that don't support 'strict-dynamic' (ignored when strict-dynamic is present) |
| base-uri | 'self' | Prevents base tag injection attacks |
| object-src | 'none' | Blocks all plugin content (Flash, Java applets) |
FOUC Prevention Script
The inline theme-detection script in head.html runs with the aem nonce, making it the trust anchor for all subsequent script loading.
HTML Sanitization
RHEA uses two sanitizers at different points. They are complementary, not alternatives — see ADR-005.
| Tool | Used in | Purpose |
|---|---|---|
| scripts/utilities/sanitize.js | footer, header, search, cookie-consent-banner, api-widget, isi | Removes the sink: inserts DOM nodes instead of assigning a string |
| DOMPurify (scripts/dompurify.min.js) | isi, embed, scripts/editor-support.js | Sanitizes a string at the boundary, where content arrives |
isi uses both: DOMPurify cleans fragment content as it is read, then setSanitizedHTML inserts it without an innerHTML assignment.
Sanitization Pattern
import { setSanitizedHTML, sanitizeToFragment } from '../../scripts/utilities/sanitize.js';
// Insert authored or fetched content -- no innerHTML sink at the call site
setSanitizedHTML(element, rawHTML);
// Or, when you need nodes rather than a string
element.append(sanitizeToFragment(rawHTML));
element.innerHTML = DOMPurify.sanitize(html) was the previous pattern and is no longer correct. Sanitizing into a sink still leaves the sink: Snyk's taint analysis reports the assignment regardless, and — more importantly — a later edit can drop the sanitizer while the assignment survives. Removing the assignment removes that failure mode entirely.
All 14 DOM XSS findings were resolved this way, taking snyk code test to 0. See ADR-012.
- Inserting authored or fetched content into the DOM -> setSanitizedHTML / sanitizeToFragment
- Cleaning a string where it arrives, in a block that already uses DOMPurify -> keep DOMPurify
- Interpolating into a quoted attribute -> escapeHTML (escapes both quote characters)
- Do not add a third sanitizer. api-widget keeps a local allow-list sanitizer and a quote-preserving escapeText for documented reasons — see ADR-013.
Regression coverage
tests/blocks/xss.spec.js in the RHEA repo pins every sanitizer sink — 11 tests, driving each block through its real decorate() entry point. It asserts on a global execution canary rather than on rendered nodes, because a detached-but-owned element still fires handlers.
npx playwright test tests/blocks/xss.spec.js
Dependency Security
Renovate Configuration
Automated dependency management via .renovaterc.json (see the RHEA source repository). Key security impacts: devDependencies auto-merge for rapid patching, continuous schedule for zero-delay vulnerability response. See the DevOps Overview for how dependency and security checks fit into CI.
Supply Chain Controls
| Control | Implementation | Description |
|---|---|---|
| Lock file | package-lock.json | Exact dependency versions pinned |
| npm ci | CI pipeline | Clean install from lock file only |
| Renovate | Automated PRs | Dependency updates with changelogs |
| Snyk scan | GitHub Actions (Snyk-Scan job on pull requests) | Third-party vulnerability scanning via the shared abbvie-migrated/SnykTemplates workflow |
| Limited runtime deps | Architecture choice | Both runtime libraries (mitt, DOMPurify) are vendored/bundled locally — no third-party CDN runtime dependency |
External Script Policy
| Script | Source | Loading Method | Trust Mechanism |
|---|---|---|---|
| mitt (event emitter) | Local bundle (scripts/vendor/mitt.min.js) | ES module import | Same-origin, nonce-trusted |
| DOMPurify | Local bundle (scripts/dompurify.min.js) | ES module import | Same-origin, nonce-trusted |
| AEM RUM | ot.aem.live | Dynamic script tag | CSP strict-dynamic |
| Block JS | Same-origin | Dynamic import() | Same-origin, nonce-trusted |
Access Control
Repository Access
| Level | Scope | Description |
|---|---|---|
| Admin | abbvie-migrated org admins | Full repository control |
| Write | RHEA development team | Push, merge, branch management |
| Read | Consuming site teams | Clone, pull, fork |
AEM Author Access
| Role | Permissions | Description |
|---|---|---|
| Content Author | Create, edit, publish pages | Day-to-day content management |
| Content Admin | Fragment management, site config | Global component management |
| Developer | All above + code deployment | Full authoring and development access |
Security Best Practices
For Block Developers
| Practice | Description |
|---|---|
| Insert external HTML without a sink | Use setSanitizedHTML() / sanitizeToFragment() for content from fragments, APIs, or editor events. Reach for DOMPurify only at a boundary where a string arrives — see HTML Sanitization |
| Never assign innerHTML, sanitized or not | Sanitizing into a sink leaves the sink; a later edit can drop the sanitizer while the assignment survives |
| Avoid eval() and new Function() | CSP will block these |
| Use loadScript() for dynamic scripts | Maintains CSP trust chain via strict-dynamic |
| Validate user input | Client-side validation for form fields (server-side is the ultimate authority) |
| Use getLinkAttributes() | Automatically adds rel="noopener noreferrer" for external links |
| Avoid inline event handlers | Use addEventListener() instead of onclick attributes |
For Content Authors
| Practice | Description |
|---|---|
| Review ISI content | Legal review before publishing safety information |
| Verify external links | Ensure links point to trusted destinations |
| Use fragment system | Load shared content via fragments, not inline HTML |
Vulnerability Response
| Step | Action | Owner |
|---|---|---|
| 1 | Detection — Renovate PR, npm audit, or manual report | Automated / Any team member |
| 2 | Assessment — Evaluate severity and exploitability | Security lead |
| 3 | Patch — Update dependency or apply code fix | Developer |
| 4 | Verify — Confirm vulnerability is resolved | Developer |
| 5 | Deploy — Merge fix to main | Developer |
Related Documents
- DevOps Overview — Security gates in the pipeline
- Support and Incident Response — Incident response procedures
- Architecture Decisions — Security-related architecture decisions