api/documentation/security/pen-test-plan.md

# White-Hat Penetration Test — Plan & Engagement (TRU-80)

> ## ⛔ CANCELLED — DO NOT EXECUTE
> **Status:** Cancelled 2026-08-16 by Jeppe Bundgaard
> **Reason:** No budget approved at this time. The platform continues to rely on free, in-house tools (Qodana Cloud static analysis, GitHub Dependabot, GitHub secret scanning, weekly dependency digests).
> **What this means:** No external pen-test firm is being engaged. This document is kept as a planning artifact for future reference. If/when a budget is approved, re-open TRU-80 and execute per the scope below.
> **Owner:** Jeppe Bundgaard (jeppe@copenhagentruckwash.io)
>
> ---

**Linear:** [TRU-80 — DRIFT 19: White hat pen test (security review)](https://linear.app/truck-wash-aps/issue/TRU-80/drift-19-white-hat-pen-test-security-review)
**Project:** UI Library & Pen Testing
**Priority:** Medium
**Status (this doc):** Draft v1 — ready for engineering + management review
**Author:** bugfix sub-agent (TRU-80)
**Date:** 2026-08-16

---

## 1. Purpose

Define the scope, methodology, deliverables, scheduling, and budget envelope for an
independent white-hat penetration test of the Truck Wash ApS platform. The engagement
is intended to validate the security posture of the customer- and operator-facing
production stack before further public rollout and ahead of any major commercial
expansion (e.g. additional self-serve sites, additional payment integrations).

This document is the planning artefact for TRU-80. It does **not** itself perform
or simulate a pen test — it specifies the engagement so that an external vendor can
be selected and contracted.

---

## 2. Scope (in)

The following systems are **in scope** for the engagement. Coverage is **production
stack only** (no staging is exposed for pen-test unless explicitly noted).

### 2.1 API (PHP / NGINX, `copenhagentruckwash/api`)

- All HTTP(S) routes under `services/nginx/app/routes/` (≈116 route files) and
  `services/nginx/app/modules/*/routes/` (multiple modules incl. Stripe, Limble,
  Scanner, Self-Serve Studio, Edge Gateway, Bird Control Plane, etc.).
- Authentication / session endpoints, including:
  - `usersRoute.php`, `userSecurityRoute.php`, `superuserSecurityRoute.php`,
    `subusersRoute.php`, `limitedBackofficeRoute.php`
  - `limitedBackofficeLoginGrantService.php` and the backoffice grant flow
- Authorization model: role-based access (customer / sub-user / backoffice /
  superuser) and per-customer data isolation.
- Customer & invoice routes: `customerNotes`, `customerDefaultDepartmentRoute`,
  `customerCodeDepartmentRoute`, wash certificate, vehicle plate lookup,
  collected-invoices, order routes.
- Payment integration: Stripe module (`moduleStripeRoute.php`).
- Economic ERP integration (`economic_endpoint_t.php` trait) — read-only
  token handling, invoice push.
- Edge gateway / IoT surface: `moduleEdgeGatewayRoute.php`, `edgegateway.php`,
  `shelly.php`, `gateway_shelly_transport.php`, `birdControlPlaneRoute.php`.
- File / media endpoints: `file_server.php` (auth-gated downloads, S3 / local).
- Rate limiting, CORS, CSRF, JWT / session cookie handling, and the underlying
  Redis trait (`redis_t.php`).
- WordPress trait / integration (`wordpress_api_object_t.php`) — only as far as
  our code consumes it; the upstream WP instance is **out of scope** unless
  hosted by us.
- Container/infrastructure: `Dockerfile`, `Dockerfile.coolify-api`, NGINX
  config (`nginx.conf`, `apache-ssl.conf`), `docker-compose.prod.yml`,
  `coolify` deploy config. Black-box reachable attack surface only.

### 2.2 Pleno-Vue (Vue 3 + Capacitor, `copenhagentruckwash/pleno-vue`)

- Web SPA (`app/`, `index.html`, `dist/`) reachable at the production hostname.
- Mobile builds for Android (`android/`, `build.gradle`, `fastlane/`) and iOS
  (`ios/`) packaged via Capacitor (`capacitor.config.ts`).
- API client and token storage in the SPA (where tokens live, at-rest
  protection, refresh flow).
- Build-time secrets, env handling (`env.d.ts`, `manifest-checksum.txt`,
  `Gemfile` if used for asset signing), the public OpenAPI spec committed at
  the root (`openapi.yaml`).
- Capacitor deep-link / universal-link / custom-scheme handling
  (`capacitor.config.ts`).

### 2.3 Infrastructure & cross-cutting (in)

- TLS configuration (cert chain, HSTS, cipher suites) on the production
  public host.
- HTTP security headers (CSP, X-Frame-Options, Referrer-Policy,
  Permissions-Policy, X-Content-Type-Options).
- Subdomain / wildcard exposure (`*.truckwash.dk` style).
- Email & SMS notification paths only as far as they can be abused for
  spoofing / phishing of our users (we control the From domain).

### 2.4 Out of scope (explicitly)

- Upstream SaaS providers' own infrastructure: Stripe, Economic, WordPress.com,
  Shelly cloud, Limble, Mailgun, etc. We will only test the **integration**,
  not the third party itself.
- Internal office LAN, employee laptops, MDT, and physical site hardware
  (gate controllers, scanners) — these are covered by a separate physical /
  OT scope and **out of scope** for this IT pen test.
- Denial-of-service / load testing.
- Social engineering of Truck Wash staff.
- Source-code review of `node_modules` / vendor dependencies (the engagement
  will use SCA tooling to flag known CVEs, but not audit transitive deps).
- Any production data exfiltration — the vendor will be given sanitised or
  test accounts and synthetic data only.

---

## 3. Methodology

Industry-standard, manual-led engagement with tooling support. Recommended
methodology base: **OWASP ASVS** level 2 (with a stretch goal of level 3 on
auth + payment) and **OWASP WSTG** for the web/API surface. Mobile builds will
use **OWASP MASVS** as the checklist.

Phases (estimated total: 12 working days of vendor effort, see §6):

1. **Scoping & recon (1 day)**
   - Confirm target list, accounts, and rules of engagement.
   - Passive recon (DNS, cert transparency, subdomains, public OpenAPI spec).
   - Active recon limited to non-destructive fingerprinting.
2. **API pen test (3 days)**
   - AuthN/AuthZ boundary testing on every route group in §2.1.
   - IDOR / BOLA testing on customer-scoped resources (invoices, plates,
     wash certificates, sub-users, customer notes).
   - Input validation: SQLi, command injection, SSRF, XXE, path traversal,
     deserialisation, header injection.
   - Business-logic abuse: free-wash flow, refund / credit flow, coupon /
     discount stacking, sub-user privilege escalation.
   - Webhook signature validation (Stripe, Edge Gateway, Shelly).
3. **Web SPA pen test (2 days)**
   - XSS (reflected, stored, DOM-based) including Vue template injection.
   - Token storage, leakage via 3rd-party scripts, postMessage abuse.
   - Open-redirect / OAuth misconfig in any SSO flow.
   - CSP / SRI effectiveness.
4. **Mobile (Capacitor) review (2 days)**
   - Static analysis of the built APK / IPA (Capacitor WebView).
   - Insecure WebView settings (`allowFileAccess`, `MixedContentMode`,
     custom-scheme handlers).
   - Local storage of tokens, biometric bypass if implemented.
   - Deep-link / universal-link hijack attempts.
5. **Infrastructure & config (1.5 days)**
   - TLS, headers, cookie flags, HSTS preload eligibility.
   - NGINX hardening review (based on provided config snapshots).
   - Docker / coolify surface only as externally reachable.
6. **SCA / dependency check (0.5 day)**
   - `composer.json` and `package.json` SCA scan.
   - High-severity known-CVE report only; no deep audit.
7. **Exploitation & PoC (1 day)**
   - Build proofs-of-concept for any Critical / High findings.
8. **Reporting & re-test (1 day)**
   - Draft report → vendor walkthrough → final report.
   - Re-test of fixed findings is scoped separately (see §6).

---

## 4. Rules of engagement (RoE)

- **Window:** business hours Europe/Copenhagen by default; out-of-hours
  exploitation only with prior written approval per critical finding.
- **Contact channel:** shared Signal thread + email; vendor given a Slack
  guest account in a dedicated `#sec-pentest-2026Q4` channel.
- **Stop conditions:** any finding that risks data loss, payment integrity,
  or production gate operation → immediate stop + phone call to on-call.
- **Data handling:** vendor may only use synthetic / test data. No
  exfiltration of real customer PII. All artifacts returned or destroyed at
  end of engagement (TBD in contract).
- **Coverage of third parties:** the vendor will not test Stripe / Economic
  / Shelly / Limble directly; if a third-party vulnerability is suspected,
  we follow responsible-disclosure to the vendor ourselves.

---

## 5. Deliverables

1. **Kick-off doc** (this plan, signed off by both parties).
2. **Daily standup notes** in `#sec-pentest-2026Q4` (one paragraph + new
   findings list).
3. **Mid-engagement check-in** at end of phase 3 — informal review of any
   Critical / High so we can start patching in parallel.
4. **Final report (PDF + JSON)** including:
   - Executive summary, risk heatmap, business-impact narrative.
   - Each finding: title, CVSS v3.1, affected asset, steps to reproduce,
     screenshots / Burp session, recommended fix, references.
   - SCA dependency report as an appendix.
5. **Re-test letter** (separate SOW, see §6).
6. **Knowledge transfer**: 60-min session for engineering on the top 5
   findings.

---

## 6. Budget & scheduling

### 6.1 Indicative effort

| Phase | Days | Notes |
| --- | --- | --- |
| 1. Scoping & recon | 1.0 | joint with us |
| 2. API pen test | 3.0 | |
| 3. Web SPA | 2.0 | |
| 4. Mobile (Capacitor) | 2.0 | |
| 5. Infra & config | 1.5 | |
| 6. SCA | 0.5 | tooling-led |
| 7. Exploitation / PoC | 1.0 | |
| 8. Reporting | 1.0 | incl. 1 review round |
| **Total** | **12.0 days** | |

### 6.2 Indicative cost (DKK, ex. VAT)

Pricing varies significantly with vendor. Three realistic budget tiers for
procurement:

| Tier | Daily rate (DKK) | Total (12 d) | Notes |
| --- | --- | --- | --- |
| Boutique / Nordic boutique (e.g. Danish / Swedish) | 12 000 – 16 000 | **144 000 – 192 000** | Best fit for our stack size, Danish-language reporting available. |
| Mid-tier international (e.g. NCC, Securix, Pentest People) | 15 000 – 22 000 | **180 000 – 264 000** | More brand name, more bureaucracy, stronger report templates. |
| Top-tier / Big-4 style | 25 000 – 40 000 | **300 000 – 480 000** | Overkill for current footprint; revisit at Series-A. |

**Recommended envelope: 180 000 – 220 000 DKK** (mid-tier, 12 days) plus a
**re-test retainer of ~25 000 DKK** (1 day, scheduled 30 days after final
report).

Add ~5 000 DKK contingency for incident-response hours if a Critical is
found mid-engagement.

### 6.3 Schedule (proposed)

- **2026-08-25** — this plan reviewed and signed off by management.
- **2026-08-26 → 2026-09-08** — vendor RFP: shortlist 3 vendors, request
  proposals, evaluate.
- **2026-09-09 → 2026-09-15** — contract + NDA + RoE finalisation.
- **2026-09-22 (week 39)** — engagement kick-off.
- **2026-09-22 → 2026-10-07** — on-site / remote testing (2.5 calendar
  weeks, vendor working in parallel with their normal cadence).
- **2026-10-08** — draft report.
- **2026-10-15** — final report + walkthrough.
- **2026-11-15** — re-test (retainer).

All dates are **provisional** until a vendor is selected.

### 6.4 Vendor shortlist (candidates to approach)

We will request proposals from at least 3 of the following (final shortlist
to be confirmed with management):

1. **Securix** (DK) — boutique, OWASP ASVS-aligned, good fit for our size.
2. **Pentest People** (UK / EU) — mid-tier, mobile capability.
3. **NCC Group / nCC / NowSecure** (international) — heavier, good brand
   for enterprise due-diligence.
4. **Curity** (SE) — strong API / OAuth expertise, fits our auth model.
5. **Deutsche Cyber AG / similar Nordic boutique** — fallback.

Procurement will evaluate on: relevant references (Logistics / IoT / payment),
ASVS/MASVS familiarity, daily rate, lead time, report quality, re-test terms.

---

## 7. Pre-engagement hardening checklist (for engineering, run in parallel)

We should land these before the vendor starts — they reduce noise and let
the vendor focus on real issues:

- [ ] HSTS preload submitted; `Strict-Transport-Security: max-age=63072000; includeSubDomains; preload`
- [ ] CSP `default-src 'self'` baseline, no `unsafe-inline`; report-only first
- [ ] All cookies `Secure; HttpOnly; SameSite=Lax` (or `Strict` for backoffice)
- [ ] CSRF token on every state-changing route; verified for Stripe / Edge
      Gateway webhooks
- [ ] Webhook signature verification on Stripe, Shelly, Edge Gateway
- [ ] Rate-limit on auth, password reset, and OTP endpoints
- [ ] Sub-user privilege model re-verified against `subusersRoute.php`
- [ ] File-server (`file_server.php`) path-traversal tests in CI
- [ ] SCA in CI: `composer audit` and `npm audit --omit=dev` blocking
      high+ vulns
- [ ] Mobile: `allowFileAccess=false`, mixed content disabled, JS interfaces
      removed
- [ ] Secrets: no production keys in repo (`git log -S` audit)

This list is also the basis for re-test acceptance criteria.

---

## 8. Open questions for management

1. Confirm total budget cap (recommend ≤ 220 000 DKK + 25 000 retainer).
2. Confirm legal/procurement owner and contract template.
3. Confirm whether to require a Danish-language final report (recommended).
4. Confirm re-test budget is approved up-front, or per-finding.
5. Confirm we are comfortable with the 12-day estimate, or want a lighter
   6-day "API + SPA only" first pass.

---

## 9. References

- OWASP ASVS 4.0 — https://owasp.org/www-project-application-security-verification-standard/
- OWASP WSTG — https://owasp.org/www-project-web-security-testing-guide/
- OWASP MASVS — https://mas.owasp.org/MASVS/
- OWASP API Security Top 10 (2023) — https://owasp.org/API-Security/editions/2023/
- Linear project: *UI Library & Pen Testing* (`acc087b4-b8ce-40c4-bbca-077fd93513a4`)

---

*This document is a planning artefact, not the test itself. Once approved, a
separate SOW will be drafted with the selected vendor and linked from this
issue.*