Security Testing Policy

Security Testing Policy

Project / Organization: BilkoALAI Holding AS — BalkanDrop AccountingPayment SaaSApp Policy Number: POL-SEC-TEST-001 Version: 1.0 Date: 2026-02-23 Author: CTO / Security EngineerArchitect Status: Draft Reviewers: CISO, Engineering Lead, DPOCTO Next Review: 2027-02-23 Classification: Confidential

Document History

1. Purpose & Scope

Purpose: This policy defines the security testing requirements,methodology, tools, schedule,frequency, and acceptanceremediation criteriarequirements for all systems operated by ALAI Holding AS for the BilkoDrop platform.payment Bilko handles regulated financial data (tax IDs, IBAN, accounting records) across three jurisdictions.app. Security testing is mandatory,mandatory not— optional.no system goes to production without completing applicable security tests.

Regulatory basis:

• IKT-forskriften (FOR-2003-05-21-630) §§ 5-6 — ICT security verification
• DORA (EU) 2022/2554 Art. 24-25 — Digital operational resilience testing
• Finanstilsynet licensing requirements — penetration test before production launch
• GDPR Art. 32(1)(d) — regular testing of technical measures

Scope:

• All Bilkoproduction applicationsAPIs —and Express APIendpoints (apps/api/getdrop.no, api.getdrop.no),
Next.js
• All frontendAWS infrastructure (apps/web/),App databaseRunner, layerS3, (PrismaKMS, +Secrets PostgreSQL),Manager)
• All CI/CD pipelines and externalbuild systems
• All third-party integrations (SEF,BankID, HR-FISK).Sumsub, Open Banking partners)
• Developer workstations handling Confidential or Restricted data

Policy Owner: CISO ([email protected]) Operational Owner: Security team + Engineering Lead

2. Security Testing PyramidMethodology

Approach: Shift-left security — testing integrated throughout development, not bolted on at the end.

Testing layers:

graphDevelopment TDPhase
    subgraph+-- AUTOMATED["AutomatedUnit security tests — Vitest (runscurrent: 20+ security-specific tests)
    +-- SCA — npm audit (every commit, automated)
    +-- Secret scanning — detect leaked credentials (every commit)

Build / CI pipeline)"]
        SAST["SAST\nESLint Security Rules\nTypeScript strict mode\nnpm audit\nSnyk SCA"]
        UNIT["Security Unit Tests\nVitest\nRBAC matrix tests\nOrg isolation tests\nEncryption tests\nVAT accuracy tests"]
        INT["Integration Tests\nVitestPhase
    +-- Supertest\nAuthSAST flow— tests\nJWTstatic validation\nRatecode limiting tests"]
    end

    subgraph PERIODIC["Periodicanalysis (scheduled)"]every DAST["DAST\nOWASPPR ZAP\nMonthly— planned Phase 2)
    +-- pre-release"]SCA E2E["Security— E2E\nPlaywright\nCross-tenantdependency boundaryvulnerability tests\nPrivilege escalation tests"]
    end

    subgraph MANUAL["Manualcheck (scheduled)"]every PENTEST["PR)
    +-- Secret scanning — full scan (every PR)

Deployment Phase
    +-- DAST — OWASP ZAP against staging (planned — Phase 3)
    +-- API security scan — endpoint fuzzing (planned — Phase 3)

Operational Phase
    +-- Vulnerability assessment — external attack surface (quarterly)
    +-- Penetration Test\nExternaltest vendor\nAnnual"]— REVIEW["Securitymanual Codeexpert Review\nPre-mergetesting (security-sensitivebefore PRs)\nArchitecturePhase 3 launch)
    +-- Post-incident review quarterly"]— endregression UNITtests -->after INTany -->security DAST --> PENTESTincident

3. AutomatedCurrent Security TestingTest (CI/CD)Coverage

Every push to main and every pull request triggers:

3.1 StaticVitest AnalysisUnit (SAST)Security Tests

Status: Implemented — 20+ security-specific tests Source: src/drop-app/src/__tests__/

3.2 Running the Security Test Suite

# Run all security tests
cd src/drop-app && npx vitest run --reporter verbose

# Run specific security test file
npx vitest run src/__tests__/auth.test.ts

# Run with coverage
npx vitest run --coverage

Blocking criteria: All security tests MUST pass before merge to main branch.

4. Testing Types, Tools & Schedule

4.1 SCA including— licenseSoftware complianceComposition Analysis

Status: Implemented (npm audit)

3.2

Current Securitydependency Unitsecurity Testsstatus (Vitest)

Location: apps/api/src/__tests__/security/audit):

Required

RBAC

describe('Org isolationonly — no cross-tenant data leak', () => {
  let org1Token: string
  let org2InvoiceId: string

  beforeAll(async () => {
    // Setup two orgs with data
    org1Token = signTestJWT({ org: 'org-1', role: 'owner' })
    const org2Token = signTestJWT({ org: 'org-2', role: 'owner' })

    // Create invoice in org-2
    const res = await request(app)
      .post('/api/invoices')
      .set('Authorization', `Bearer ${org2Token}`)
      .send(validInvoicePayload)
    org2InvoiceId = res.body.id
  })

  test('org-1 cannot read org-2 invoice', async () => {
    const res = await request(app)
      .get(`/api/invoices/${org2InvoiceId}`)
      .set('Authorization', `Bearer ${org1Token}`)
    expect(res.status).toBe(404) // NOT 403 — don't reveal existence
  })

  test('org-1 list does not include org-2 data', async () => {
    const res = await request(app).get('/api/invoices').set('Authorization', `Bearer ${org1Token}`)
    const ids = res.body.data.map((i: any) => i.id)
    expect(ids).not.toContain(org2InvoiceId)
  })
})

describe('Field encryption — L4 Restricted', () => {
  test('PIB stored encrypted in DB', async () => {
    const testPIB = '123456789' // fake PIB
    // Create contact with PIB
    await request(app)
      .post('/api/contacts')
      .set('Authorization', `Bearer ${ownerToken}`)
      .send({ name: 'Test', taxId: testPIB, type: 'RS' })

    // Read raw DB value — should not be plaintext
    const raw = await prisma.$queryRaw`
      SELECT "taxId" FROM "Contact" WHERE name = 'Test'
    `
    expect(raw[0].taxId).not.toBe(testPIB)
    expect(raw[0].taxId).toMatch(/^[A-Za-z0-9+/]+=*:[A-Za-z0-9+/]+=*:[A-Za-z0-9+/]+=*$/)
    // Should be base64:base64:base64 format (iv:authTag:ciphertext)
  })

  test('decrypted PIB matches original on read', async () => {
    const res = await request(app).get('/api/contacts').set('Authorization', `Bearer ${ownerToken}`)
    const contact = res.body.data.find((c: any) => c.name === 'Test')
    expect(contact.taxId).toBe('123456789')
  })
})

describe('VAT calculation accuracy', () => {
  test('RS: VAT 20% on standard goods (NUMERIC precision)', () => {
    const net = new Decimal('100.00')
    const vatAmount = net.mul('0.20')
    const gross = net.plus(vatAmount)
    expect(vatAmount.toString()).toBe('20.00')
    expect(gross.toString()).toBe('120.00')
  })

  test('HR: VAT 25% (EUR since Jan 2024)', () => {
    const net = new Decimal('100.00')
    const gross = net.mul('1.25')
    expect(gross.toString()).toBe('125.00')
  })

  test('BA: VAT 17% (UIO standard)', () => {
    const net = new Decimal('100.00')
    const gross = net.mul('1.17')
    expect(gross.toString()).toBe('117.00')
  })

  test('No float drift on invoice totals', () => {
    // Known JS float bug: 0.1 + 0.2 !== 0.3
    const line1 = new Decimal('0.10')
    const line2 = new Decimal('0.20')
    expect(line1.plus(line2).toString()).toBe('0.30')
    // Contrast: expect(0.1 + 0.2).toBe(0.3) would FAIL
  })
})

describe('Auth — JWT security', () => {
  test('expired access token returns 401', async () => {
    const expiredToken = signTestJWT({ exp: Math.floor(Date.now() / 1000) - 1 })
    const res = await request(app)
      .get('/api/invoices')
      .set('Authorization', `Bearer ${expiredToken}`)
    expect(res.status).toBe(401)
  })

  test('tampered token returns 401', async () => {
    const validToken = signTestJWT({ role: 'viewer' })
    // Tamper: change role claim in payload
    const parts = validToken.split('.')
    const payload = JSON.parse(Buffer.from(parts[1], 'base64url').toString())
    payload.role = 'owner' // attempt privilege escalation
    parts[1] = Buffer.from(JSON.stringify(payload)).toString('base64url')
    const tamperedToken = parts.join('.')
    const res = await request(app)
      .delete('/api/invoices/any-id')
      .set('Authorization', `Bearer ${tamperedToken}`)
    expect(res.status).toBe(401)
  })

  test('rate limiting: 6th auth attempt in 15min returns 429', async () => {
    for (let i = 0; i < 5; i++) {
      await request(app).post('/api/auth/login').send({ email: 'x', password: 'wrong' })
    }
    const res = await request(app).post('/api/auth/login').send({ email: 'x', password: 'wrong' })
    expect(res.status).toBe(429)
  })
})

# .github/workflows/security.yml
- name: Audit dependencies
  run: npm audit --audit-level=high
  # HIGH or CRITICALknown CVEs

Source: name: Check for secrets in code uses: trufflesecurity/trufflehog@main # Scans for committed credentials, API keys ~/ALAI/products/Drop/security/drop-security-rapport.md

Dependency update policy:

• CRITICALSecurity CVE:patches patch(Critical/High): Merge within 24SLA hours(see §5)
• HIGHMinor CVE:security patchupdates: Merge within 714 days
• MEDIUMMajor CVE:updates: patchPlanned migration within 3090 days

Approved licenses: MIT, Apache 2.0, BSD (2-clause, 3-clause), ISC

4.2 SAST — Static Application Security Testing

Status: Planned (Phase 2)

SAST rules in scope (when implemented):

• LOWSQL CVE:injection patchdetection at(parameterized nextqueries sprintvalidation)
boundary
• JWT algorithm confusion (alg: none detection)
• Hardcoded secrets / credentials
• Insecure cryptography (MD5, SHA-1, DES, RC4)
• Path traversal / LFI
• IDOR patterns (database queries without user_id scoping)
• Missing authentication middleware

4.3 DAST — Dynamic Application Security Testing (DAST)

OWASP ZAP

Schedule:Status: MonthlyPlanned (Phase 3 — before production launch)

Scope

OutDAST ofscan scope:

• SEFAll portal,24 FINAAPI portalendpoints (externalauthenticated systems)+ unauthenticated)
• RailwayAuthentication infrastructureflows (login, registration, logout, session handling)
• CloudflareRate WAFlimiting (managedverify by10/60s Cloudflare)limits enforced)
• CSRF protection (Origin header validation)
• Input validation (IBAN, currency, language, amount)
• Security headers presence (HSTS, CSP, X-Frame-Options, etc.)
• SSL/TLS configuration (TLS 1.3, no TLS 1.0/1.1)
• JWT handling (expiry, algorithm, revocation)
• Feature flag enforcement (cards endpoints return 404 when disabled)

4.4 Penetration Testing

ZAPStatus: Configuration:Required before Phase 3 production launch (Finanstilsynet licensing requirement)

Penetration test scope:

# zap-baseline.yaml
env:
  contexts:In-scope:
- name:Staging Bilko API
      urls:
        -pre-launch: https://staging.bilko.io/api/getdrop.no
authentication:- method:Production script(after #launch): ZAPhttps://getdrop.no, scripthttps://api.getdrop.no
to- authenticateAuthentication flows (login, registration, BankID OIDC callback — Phase 2)
- All 24 authenticated API endpoints
- Session management (JWT, cookie handling, revocation)
- Rate limiting and getCSRF JWT
  rules:protection
- id:AWS 10202App #Runner Absenceconfiguration
- Cloudflare WAF configuration

Out-of-scope:
- Denial of Anti-CSRFservice tokens — note (cookies are httpOnly)
      threshold: LOWattacks
- id:Social 10096engineering #of Timestampstaff Disclosurewithout —prior ignoreapproval
- AWS infrastructure itself (timestampsAWS areglobal public)responsibility)
threshold:- OFFBankID Norge AS systems
- Sumsub systems

Rules of engagement:
- Testing window: To be agreed in writing with provider
- Notify before testing: [email protected]
- Halt on critical finding: Immediately notify [email protected] + CTO

RequiredKey ZAPtest findingsareas thresholdfor (before release):Drop:

1. CRITICALJWT /manipulation HIGH:(algorithm 0confusion, allowedexpiry bypass, forged tokens)
2. MEDIUM:Session mustrevocation bebypass assessed(sessions —table known acceptable risks documentedmanipulation)
3. LOWIDOR /attacks INFORMATIONAL:(user_id documentscoping andvalidation prioritizeacross all 24 endpoints)

5. PenetrationVulnerability Testing

Frequency: Annual (or after significant architecture change) Provider: External certified pentest firm (OSCP/CREST certified)

Scope:

• Web application (app.bilko.io)
• API endpoints
• AuthenticationClassification & session management
• Multi-tenant isolation (primary focus — org isolation must be tested)
• Business logic flaws (VAT calculation, invoice numbering)
• Third-party integrations (SEF API, HR-FISK)

Rules of engagement:

• Staging environment only — no production testing without explicit CEO approval
• No DoS / DDoS testing
• No social engineering of employees
• Penetration test agreement signed before engagement begins

Remediation SLAs

5.1 Severity Classification (post-pentestCVSS findings):

v3.1)

5.2 Remediation SLAs

SLA tracking: GitHub Issues with label security-finding + severity label.

6. Security Code Review Checklist

Required for every PR touching authentication, authorization, payment processing, PII, or cryptography:

Authentication & Authorization:

•  No hardcoded credentials, JWT secrets, or API keys
•  Password hashing uses bcrypt (cost >= 12) — SHA-256 explicitly rejected (fix C4)
•  JWT validation: signature, expiry, iat claim verified via jose
•  Session revocation: all protected endpoints check sessions table for revoked = 0
•  No role flags from user-supplied input
•  All protected endpoints have authentication middleware applied
•  Logout revokes sessions server-side (not just clears cookie)

Input Validation:

•  All user inputs validated using Drop validators (validateAmount, validateIBAN, validateCurrency, validateLanguage, sanitizeText)
•  All database operations use parameterized ? queries — no string concatenation
•  User data queries include AND user_id = ? scoping (IDOR prevention)
•  No eval(), exec(), or shell command interpolation with user input

Cryptography:

•  No MD5, SHA-1, DES, RC4 (per data-encryption-policy.md §2.2)
•  Random values use crypto.randomBytes() — not Math.random()
•  IVs/nonces are random (96-bit minimum) and never reused
•  No cryptographic keys in source code, logs, or error messages

Error Handling:

•  No stack traces or internal paths in API error responses
•  Generic messages for authentication failures (no user enumeration)
•  Errors logged to Sentry internally but not exposed in API responses

Data Handling:

•  Foedselsnummer not logged in plaintext (never in Sentry, BetterStack, console.log)
•  Sensitive data not in query parameters (use POST body)
•  Feature flags checked before any cards-related endpoint
•  AML retention: user data deletion respects 5-year retention (Hvitvaskingsloven § 30)

Dependencies:

•  New dependencies reviewed with npm audit before adding
•  No end-of-life packages introduced
•  License compatible (MIT, Apache 2.0, BSD, ISC)

7. Security Testing in CI/CD Pipeline

Current pipeline (MVP):

Developer Commit
    +-- Pre-commit hook: npm audit (SCA)
        |
        v
    Git Push -> PR
        |
        v
    Vitest security tests (20+ tests)
        |
        v (all pass)
    Merge to main
        |
        v
    Deploy to AWS App Runner

Target pipeline (Phase 2+):

flowchart LR
    COMMIT[Developer\nCommit] -->|Pre-commit| SCA_INC[npm audit\nSCA check]
    SCA_INC -->|Pass| PUSH[Git Push]
    PUSH --> PR[Pull Request]

    PR --> VITEST[Vitest\nSecurity Tests]
    PR --> SAST[SAST\nSemgrep/CodeQL]
    PR --> SCA_FULL[npm audit\nFull scan]
    PR --> SECRET[Secret\nScanning]

    VITEST & SAST & SCA_FULL & SECRET -->|All pass| BUILD[Build\nDrop App]
    BUILD --> STAGING[Deploy to\nStaging]
    STAGING --> DAST[OWASP ZAP\nDAST Scan]
    DAST -->|Pass| GATE{Security\nGate}

    GATE -->|Pass| PROD[Deploy to\nProduction]
    GATE -->|Fail| BLOCK[Block Deploy\nAlert [email protected]]

Pipeline security gate criteria:

6.8. Security CodeAudit ReviewHistory

Source: ~/ALAI/products/Drop/security/drop-security-rapport.md

Security Hardening Summary (2026-02-13)

Remaining items:

9. Reporting Format

9.1 Individual Finding

Finding ID: VULN-{YEAR}-{SEQUENCE}
Title: {SHORT_DESCRIPTION}
Severity: Critical / High / Medium / Low
CVSS Score: {SCORE} (v3.1)

Description:
{DETAILED_DESCRIPTION}

Affected Endpoint / File:
- {ENDPOINT_OR_FILE}: {URL_OR_LINE_NUMBER}

Proof of Concept:
{STEPS_TO_REPRODUCE}

Impact:
{WHAT_AN_ATTACKER_CAN_DO}

GDPR/AML Implications:
{IF_PERSONAL_DATA_OR_FINANCIAL_DATA_AFFECTED}

Remediation:
{SPECIFIC_FIX}

Owner: {ASSIGNED_ENGINEER}
SLA Due Date: {DATE}
Status: Open / In Progress / Resolved / Accepted Risk

10. Metrics & KPIs

11. Bug Bounty Program

WhenStatus: requiredPlanned (mandatoryPhase pre-merge)3 — post-launch) Platform: Intigriti or HackerOne

Responsible disclosure (interim — before bug bounty): Report security vulnerabilities to [email protected].

• ChangesAcknowledgment: towithin authentication1 orbusiness authorization codeday
• Changes to encryption utilities (encryptField, decryptField)
• Changes to Prisma query patterns (potential org isolation bypass)
• New external API integrations (SEF, FINA, etc.)
• Changes to RBAC middleware or permission matrices

Who reviews: CTO or designated Senior Engineer with security background.

Checklist for security-sensitive PRs:

•  No secrets or credentials in code or config
•  All new Prisma queries include organizationId in WHERE clause
•  New endpoints have RBAC decorator applied
•  New user inputs validated with Zod schema
•  L4 Restricted fields encrypted before write, decrypted after read
•  LoggedAction entry created for all write operations
•  Rate limiting applied to new auth-adjacent endpoints

7. CI/CD Security Gates

flowchart LR
    PR["Pull Request"] --> LINT["ESLint Security\nRules"]
    LINT -->|"PASS"| AUDIT["npm audit\n--audit-level=high"]
    AUDIT -->|"PASS"| TEST["Vitest\nSecurity Test Suite"]
    TEST -->|"PASS"| SECRETS["TruffleHog\nSecret Scan"]
    SECRETS -->|"PASS"| MERGE["Merge Allowed"]
    LINT -->|"FAIL"| BLOCK["PR Blocked"]
    AUDIT -->|"FAIL"| BLOCK
    TEST -->|"FAIL"| BLOCK
    SECRETS -->|"FAIL"| BLOCK

Non-negotiable gates (cannot be bypassed with --no-verify or --force):

1. ESLint security rules: zero error findings
2. npm audit: zero HIGH/CRITICAL CVEs
3. Vitest security tests: 100% pass — especially org isolation and RBAC tests
4. TypeScript strict: zero type errors

8. Vulnerability Disclosure

Process:

1. Security researchers may report vulnerabilities to [email protected]
2. Acknowledgment within 24 hours
3. Investigation and severity assessmentTriage: within 5 business days
4. RemediationSafe perharbor: SLAresearchers acting in Sectiongood 5faith are protected from legal action
5. Responsible disclosure: researcher notified when fix is deployed

Approval