Architecture Security Mapper | S23k

Architecture Security Map

European firewalls don't solve an American problem.

Organizations buy European alternatives. Think the problem is solved.

The problem is architecture.

  • Your primary firewall is American
  • Your failover is also American
  • Cloud Act request comes in
  • Both offline
  • Now what?

Or:

  • You buy a European firewall
  • That runs on American cloud
  • With American DNS
  • Managed via American SaaS
  • Still vulnerable

Security is not determined by the logo on your invoice.
Security is determined by what happens under pressure.

This tool shows where your architecture makes you vulnerable.
Not theoretically. Realistically.

7 components. Concrete scenarios. Architecture-based solutions.

The Path To Architectural Sovereignty

You won't be sovereign overnight. But you can take steps.

1️⃣
Map It Out
Which processes are critical to your business?
→
2️⃣
Follow The Data
Where do those datastreams flow through?
→
3️⃣
Secure Failover
Build architecture that keeps working under pressure

This is a path. Not a checklist.

Component 1: DNS Infrastructure

DNS: The Invisible Choke Point

Nobody thinks about DNS until it's too late.
But DNS sees everything. Can stop everything. Determines where your traffic goes.

Typical DNS Exposure
PRIMARY DNS
🇺🇸 US Provider
+
FAILOVER DNS
🇺🇸 US Provider
=
Single Point
of Failure
▼
Architecture-Based Solution
PRIMARY DNS
🇺🇸 US Provider
+
FAILOVER DNS
🇪🇺 EU Provider
=
Keeps Working
Under Pressure

Realistic Failure Scenario

  • Day 0: Cloud Act request at your DNS provider. They must log your traffic.
  • Day 1: Subpoena. They must hand over your zone files.
  • Day 2: Regulatory order. They must block certain records.
  • Day 3: You want to migrate. Contract clause says 30 days notice.
  • Day 4: You try zone transfer. API access blocked "pending legal review".

Question: What percentage of your organization is now offline?

Architecture Reality Check

  • "We hebben backup DNS" → Bij dezelfde provider? Zelfde jurisdictie?
  • "We gebruiken Europese provider" → Maar Amerikaanse upstream/parent company?
  • "We hebben monitoring" → Die je vertelt dat DNS down is, maar niet hoe te herstellen
  • Primary + Secondary in verschillende jurisdicties → Amerikaanse primary, Europese secondary die binnen 60 seconden primary kan worden
  • Zone file autonomie → Je hebt altijd actuele zone files buiten provider control
  • Recursive resolver control → Eigen resolvers die je in uren kan omrouten
  • Multi-registry strategie → Niet alle domeinen via één registrar
Score being calculated...
📚 Deep Dive: DNS Implementation ▼

Concrete Implementation Steps

1. Zone File Autonomie

# Automated zone file backup (daily cron) 0 2 * * * /usr/local/bin/dns-backup.sh # dns-backup.sh exports zones naar off-site locatie # Niet bij DNS provider, maar eigen storage # Format: BIND zone file (universeel compatibel)

2. Multi-Jurisdictie Setup

  • Primary NS: Amerikaanse provider (snelheid, DDoS protection)
  • Secondary NS: Europese provider (jurisdictional diversity)
  • Hidden primary: On-premise (ultimate control)
  • DNSSEC: Signed zones prevent tampering

3. Tested Failover Procedure

# Quarterly drill scenario: # T+0: Primary DNS provider offline # T+60sec: Secondary promoted via registrar # T+5min: Verify propagation globally # T+24hr: Primary restored, failback tested

ZAA Principes:

  • Assume Compromise: Elke NS kan morgen compromised zijn
  • Design for Failure: Kan je binnen 60 sec omschakelen?
  • Verify Independence: Test quarterly of failover echt werkt
→ Read full ZAA Doctrine

Component 2: Authentication & Identity

Auth: The Keys To The Kingdom

Your entire security model depends on one question:
Who can impersonate your users?

Auth Dependency Chain
10,000 Users
→
500 Applications
↓
1 Amerikaanse IdP
Vendor-managed recovery | Cloud-hosted tokens
↓
Under Cloud Act request:
→ All auth logs handed over
→ Recovery flows blocked
→ Complete business standstill
Solution: Federated identity + break-glass procedures

Realistic Failure Scenario

  • Day 0: Your American IdP receives Cloud Act request. They must hand over all auth logs.
  • Day 1: You discover logs contain all user movements. Who logged in when and where. To what.
  • Day 2: Regulatory panic. You must explain why American service has access to user data.
  • Day 3: You decide to migrate. Then you discover: no user export with password hashes.
  • Day 4: You realize: all API tokens, OAuth flows, SAML connections → tied to that IdP.
  • Day 5: Migration means: reconfigure 500 applications. Re-onboard 10,000 users.

Question: How much business downtime can you survive?

Architecture Reality Check

  • "We hebben SSO" → Via één Amerikaanse provider = single point of failure
  • "We hebben MFA" → Controlled by vendor die het kan uitschakelen
  • "We hebben backup admin accounts" → Die ook via die IdP lopen
  • Federated identity met secundair systeem → LDAP/AD dat binnen uren primary kan worden
  • Break-glass procedures → Lokale admin accounts die werken als IdP down is
  • Token autonomie → JWT signing keys onder eigen beheer
  • User directory ownership → Continuous sync naar eigen database
  • Staged migration capability → Parallel auth systems tijdens transitie
Score being calculated...
📚 Deep Dive: Auth Architecture ▼

Break-Glass Procedures

Scenario: Primary IdP Offline

# Emergency auth activation 1. Key custodian A + B meet (quorum) 2. Unlock break-glass credentials (vault) 3. Activate secondary LDAP/AD 4. Update application auth endpoints 5. Notify users via alt channel (SMS) 6. Time to operational: <4 hours

Federated Identity Setup

  • Primary: Cloud IdP (convenience, features)
  • Secondary: On-prem AD/LDAP (always available)
  • Continuous sync: User directory naar beide
  • Break-glass: Local admin accounts (not via IdP)

Token Sovereignty

# JWT signing keys onder eigen beheer # Niet vendor-managed, maar HSM-backed # Key rotation: automated, no vendor dependency # Token validation: kan on-prem if IdP down

Testing Protocol:

  • Quarterly: Simulate IdP outage (test environment)
  • Measure: Time to activate secondary auth
  • Validate: Can users actually login via secondary?
  • Document: What broke? What worked?
→ Reversibility Playbook: Identity Providers

Component 3: Firewall & Network Security

Firewall: The Example Nobody Thinks Through

"We'll buy a European firewall."
Fine. And your failover? Also European?
And the management interface? Where does that run?
And firmware updates? Where are those pushed from?

One weak link. Your entire network open.

Attack Surface Analysis
🔥
Primary Firewall
US Vendor
🔥
Failover Firewall
Same Vendor
☁️
Management
Cloud SaaS
⬇️
Updates
Auto-pushed
4 Attack Vectors → 1 Vendor
Under compromise: complete perimeter open

Realistic Failure Scenario

  • Day 0: Cloud Act request at your firewall vendor. They must grant management access.
  • Day 1: Via management interface a "security update" is pushed. Actually a backdoor.
  • Day 2: Your IDS detects strange traffic from firewall to American IPs. Encrypted.
  • Day 3: You want to isolate firewall. Then realize: no internet traffic = business stopped.
  • Day 4: You activate failover. Turns out: same vendor. Same management plane. Same backdoor.

Question: How many days can you go without perimeter security?

Architecture Reality Check

  • "We hebben redundante firewalls" → Van zelfde vendor = zelfde exposure
  • "We hebben Europees product" → Maar Amerikaanse parent company of management cloud
  • Primary/Failover jurisdictional diversity → Verschillende vendors, verschillende jurisdicties
  • Management plane separation → Lokale management, geen cloud dependencies
  • Config-as-code autonomie → Ruleset in Git, deploybaar naar elk systeem
  • Regular failover drills → Quarterly test van failover procedures
Score being calculated...
📚 Deep Dive: Firewall Diversity ▼

Defense in Depth = Defense in Diversity

Architecture Principle:

Redundancy with same vendor = illusion of safety. True resilience comes from fundamentally different technologies.

# Example Stack: Primary: Palo Alto (feature-rich, cloud-managed) Failover: pfSense (open-source, on-prem managed) Tertiary: OPNsense (different codebase, diverse) # Ruleset: Infrastructure as Code - Git repository met firewall rules - Deployable naar elk platform binnen uren - No vendor lock-in op rule syntax

Config-as-Code Example:

# Terraform/Ansible managed rules # Abstract representation → platform-specific rule "allow_https" { source: internal_network dest: internet port: 443 action: allow } # Deployt naar: Palo Alto, pfSense, iptables # Zelfde rule, verschillende implementaties

Failover Drill Checklist:

  • Primary offline → Traffic via failover binnen 10 sec
  • Alle rules actief? VPN tunnels up?
  • Logging still functional? SIEM seeing traffic?
  • Kan je ruleset updaten via failover management?
  • Measure: downtime, packet loss, recovery time

Why this works: Als Palo Alto firmware compromised wordt, heeft aanvaller geen access tot pfSense. Verschillende codebase, verschillende vulnerabilities, verschillende management planes.

Component 4: Logging & SIEM

Logging: The Archive That Betrays You

Logs are your audit trail. Your security evidence. Your regulatory must-have.
They're also a complete reconstruction of what happens in your organization.

Question: wie leest mee?

Realistic Failure Scenario

  • Day 0: Cloud Act subpoena at your SIEM provider. They must hand over 90 days of logs.
  • Day 1: Those logs contain: user logins, access patterns, API calls, database queries, email metadata.
  • Day 2: You realize: this is a complete reconstruction of who-what-when in your organization.
  • Day 3: Legal team asks: "can we delete those logs?" Answer: No. Contract says 90 days retention.
  • Day 4: You want to migrate to different SIEM. Discover: queries are vendor-specific. Dashboards are vendor-specific.

Question: How much operational intelligence are you giving away?

Architecture Reality Check

  • "Logs zijn encrypted" → Bij vendor die de keys heeft
  • "We gebruiken Europese region" → Van Amerikaanse vendor onder Amerikaanse law
  • Time-based tiering → Real-time analyse in SaaS (7 dagen), long-term storage on-prem
  • Log format sovereignty → Standard formats (Syslog, JSON) niet vendor-proprietary
  • Duplicate streaming → Logs naar SaaS én lokaal systeem simultaneously
  • Hot-swap capability → Binnen 48 uur van SaaS naar on-prem
Score being calculated...
📚 Deep Dive: Log Sovereignty ▼

Time-Based Tiering Strategy

# Log flow architecture: Day 0-7: SaaS SIEM (real-time analysis) Day 8-90: On-prem storage (warm tier) Day 90+: Cold archive (compliance) # Duplicate streaming: All logs → SaaS SIEM (for analysis) All logs → Local syslog (for sovereignty)

Log Format Sovereignty:

  • Standard formats: Syslog, JSON, CEF
  • No vendor-proprietary encoding
  • Parsing rules in code (not vendor UI)
  • Detection logic portable across platforms

Hot-Swap Capability:

# Scenario: Must leave SaaS SIEM Day 0: Decision made Day 1: Spin up on-prem ELK/Splunk Day 2: Duplicate log streams activated Day 3-30: Run parallel (validate detection) Day 31: Cut over to on-prem primary Day 32: Decomission SaaS # Requirements: - Log shippers not SaaS-specific (Filebeat, Fluentd) - Detection rules in Sigma format (universal) - Dashboards in code (Grafana, Kibana)

What You Should Never Do:

  • ❌ Only store in SaaS (no local copy)
  • ❌ Proprietary log formats/encoding
  • ❌ All detection rules in vendor UI only
  • ❌ Long retention in external jurisdiction

Remember: Log data is intelligence. Treat it like classified material. Question: "Wie mag dit lezen onder welke jurisdictional pressure?"

Component 5: Encryption Key Management

Keys: The Ultimate Control

Encryption is worthless if someone else has the keys.
And "Hardware Security Module" means nothing if that HSM is managed via American cloud.

Who can decrypt your data? Really?

Key Control Hierarchy
☁️ Cloud Provider Master Key
Vendor has ultimate control
↓
🔑 Your "Encryption Keys"
Encrypted with vendor's master key
↓
💾 Your Data
"Encrypted at rest"
Reality: Vendor can decrypt at any moment
Master key access = game over

Realistic Failure Scenario

  • Day 0: Cloud Act order at your KMS provider: "hand over encryption keys for customer X".
  • Day 1: Provider must comply. Your keys are handed over. You don't know it.
  • Day 2: With your keys they can: decrypt your backups. Read your databases. Open your archives.
  • Day 3: You discover it through indirect evidence. You want to rotate keys. Then realize: rotation via vendor API.
  • Day 4: You try immediate re-encryption with new keys. Workload: 10TB data. Time: 72 hours.

Question: How much of your data do you still consider confidential?

Architecture Reality Check

  • "We gebruiken HSM" → Maar managed door Amerikaanse vendor
  • "Keys zijn encrypted" → Met master key van provider
  • "We hebben BYOK (Bring Your Own Key)" → Die via hun API moet uploaden
  • On-premise HSM met vendor-agnostic protocols → PKCS#11, niet proprietary API
  • Shamir Secret Sharing → Keys split across multiple custodians
  • Offline key generation → Critical keys created in air-gapped ceremony
  • Cryptographic autonomy → Generate, rotate, backup keys zonder external dependency
Score being calculated...
📚 Deep Dive: Key Ceremonies & HSM Setup ▼

Cryptographic Autonomy

Key Generation Ceremony (Airgapped):

# Root CA generation (offline, witnessed) 1. Clean room setup (no network) 2. Witnesses: 3+ key custodians 3. Generate root CA key (4096-bit RSA or ECC) 4. Split key: Shamir Secret Sharing (3-of-5) 5. Distribute shares to custodians 6. Secure root CA in vault (offline) 7. Generate intermediate CAs (online HSM) 8. Document everything, video record ceremony # Result: Root CA offline, compromising requires # physical access to 3+ custodian shares

HSM Architecture:

  • On-premise HSM (niet cloud HSM-as-a-Service)
  • PKCS#11 interface (vendor-agnostic)
  • Clustered voor redundancy (different vendors mogelijk)
  • Key backup: Encrypted with custodian keys

Emergency Re-Key Procedure:

# Scenario: HSM/keys compromised Hour 0-2: Custodians convene (quorum) Hour 2-4: Generate new root CA from vault shares Hour 4-8: Issue new intermediate CAs Hour 8-24: Reissue all certificates Hour 24-72: Re-encrypt all data (phased) # Tested: Annually # Measured: Time to complete, data downtime

BYOK is Not Enough:

Bring Your Own Key (BYOK) to cloud providers sounds good, maar:

  • Je uploadt key via vendor API
  • Key wrapped/encrypted by vendor master key
  • Vendor heeft runtime access voor operations
  • At best: security through obscurity

Critical: The phrase "we use encryption" is meaningless without answering "who controls the keys under what conditions". Encryption zonder key sovereignty is theater.

Component 6: Data Storage & Backups

Storage: Where Your Data Really Lives

"Our data is in the EU."
Nice. On whose disks? Whose hypervisor? Whose control plane?
And your backups? Same provider? Same jurisdiction?

Data sovereignty is an architecture question, not a contract clause.

Realistic Failure Scenario

  • Day 0: Cloud Act order: provider must grant full storage access.
  • Day 1: You're not informed (gag order). Compliance happens in silence.
  • Day 2: Months later: regulatory investigation. "How was this data obtained?"
  • Day 3: You discover: backups were also accessible. Deleted data? Restored from backup snapshots.
  • Day 4: You want to move data. Egress costs: €50K. Transfer time: 2 weeks.

Question: How much of your data pipeline must you rewrite?

Architecture Reality Check

  • "Data is encrypted at rest" → Met keys accessible via vendor API
  • "We hebben geografische redundantie" → Bij zelfde vendor = zelfde exposure
  • Storage diversity → Primary bij vendor A, backups bij vendor B
  • Application abstraction → S3-compatible API, swappable backends
  • Continuous offline sync → Periodic pull naar on-premise storage
  • Tested migration path → Quarterly drill: complete restore binnen 48 uur
Score being calculated...
📚 Deep Dive: Storage Portability ▼

Application Abstraction Layer

# Storage interface: S3-compatible API # Application never talks directly to AWS/Azure/GCP app → storage_abstraction_layer → actual_storage # Backend swappable: - MinIO (on-prem, S3-compatible) - Wasabi (S3-compatible, non-US) - Backblaze (S3-compatible, EU region) - AWS/Azure/GCP (when convenient) # Switch backend: update 1 config file # No application code changes needed

Continuous Offline Sync:

  • Primary: Cloud storage (performance, features)
  • Secondary: Different cloud (jurisdictional diversity)
  • Tertiary: On-premise (air-gapped, weekly sync)
  • All copies: Encrypted with keys YOU control

Migration Drill (Quarterly):

# Scenario: Provider X shutting down in 7 days Day 0: Decision to migrate Day 1: Spin up MinIO cluster (on-prem/alternate cloud) Day 2: Start data sync (parallel to production) Day 3-5: Incremental sync, validate data integrity Day 6: Final sync, update app config Day 7: Cut over, verify all systems operational # Measure: - Time to operational on new storage - Data loss (none acceptable) - Application downtime - Cost (egress, compute, re-architecture)

Egress Planning:

Cloud providers make it expensive to leave. Budget for:

  • Egress costs: €50-100 per TB is normaal
  • Transfer time: 10TB = days, 100TB = weeks
  • Parallel operation: Run old + new tijdens migratie
  • Validation: Verify data integrity post-migration

Critical Question: "If we received notice tomorrow that vendor X is shutting down, how many days until we're operational elsewhere?" If answer is >7 days, your architecture has a problem.

Component 7: Compute & Application Runtime

Compute: Where Your Code Actually Runs

Your application can be perfectly secure.
But if it runs on American hypervisor...
With American orchestration...
And American container registry...

Then your application is a guest in someone else's house.

Realistic Failure Scenario

  • Day 0: Compliance order: cloud provider must hand over VM snapshots.
  • Day 1: With those snapshots they get: your application code, config files, environment variables, credentials.
  • Day 2: You realize: forensically, a VM snapshot is more complete than source code.
  • Day 3: You want to migrate workloads. Discover: application uses 15 different cloud-specific services.
  • Day 4: Re-engineering estimate: 4 months development, 2 months testing. Cost: €800K.

Question: How much of your tech stack is actually reusable?

Architecture Reality Check

  • "We gebruiken Kubernetes" → Op vendor-managed control plane
  • "We hebben multi-region" → Bij zelfde vendor = zelfde exposure
  • Cloud-agnostic abstractions → Terraform modules that work across providers
  • Standard protocols over proprietary APIs → PostgreSQL over RDS
  • Container portability → OCI-compliant images
  • Tested portability → Regular drill: deploy to different cloud binnen 24 uur
Score being calculated...
📚 Deep Dive: Cloud-Agnostic Architecture ▼

Portability by Design

Infrastructure as Code (Multi-Cloud):

# Terraform modules: cloud-agnostic module "compute" { source = "./modules/compute" provider = var.cloud_provider # aws, azure, gcp, on-prem # Abstract resources, not cloud-specific instances = 10 cpu = 4 memory = 16 } # Deploy to AWS: terraform apply -var cloud_provider=aws # Deploy to Azure: terraform apply -var cloud_provider=azure # Same infrastructure, different backend

Standard Protocols Over Proprietary:

  • ✓ PostgreSQL (portable) vs ✗ AWS RDS Aurora (lock-in)
  • ✓ Redis (portable) vs ✗ ElastiCache proprietary features
  • ✓ Kafka (portable) vs ✗ AWS Kinesis (lock-in)
  • ✓ MinIO (S3-compatible) vs ✗ S3-specific APIs

Container Portability Checklist:

# OCI-compliant images (not vendor-specific runtime) # Standard orchestration (Kubernetes, not ECS/AKS-specific) # No cloud provider sidecars/agents embedded # Environment config via standard env vars # No hardcoded cloud-specific endpoints # Result: docker build → runs anywhere # AWS EKS → Azure AKS → On-prem K8s → no changes

Vendor Service Inventory:

Maintain a living document:

  • Every proprietary service in use
  • What it does (function)
  • Open-source equivalent (exit path)
  • Migration effort (hours/days/months)
  • Business impact with vendor loss

Portability Testing:

# Semi-annual drill: 1. Take production app 2. Deploy to completely different cloud/on-prem 3. Measure: setup time, issues encountered 4. Document: what broke, what worked 5. Fix: remove vendor dependencies discovered 6. Repeat until <24 hour deployment time # Goal: Confidence you CAN leave if needed

Architecture Principle: Every time you add a vendor-specific service, ask: "what is my exit cost?" If you cannot answer that within 30 seconds, do not add it.

Your Architecture Security Assessment

Overall Architecture Security Score
-

Jurisdiction Exposure

Under regulatory pressure, how much remains?

🔴 American Jurisdiction -
Can become inaccessible under Cloud Act request
🟡 European Jurisdiction -
More protected but check parent companies
🟢 Own Control -
Remains operational under jurisdictional pressure

Critical Choke Points

These components form your greatest risk:

Architecture Beats Compliance. Always.

Je hebt nu gezien waar je kwetsbaar bent. Not theoretically. Realistically.

These aren't hypothetical scenarios. These are things that happen:

  • Cloud Act requests come in. Daily.
  • Vendors change ownership. Quarterly.
  • Control planes get compromised. Weekly.
  • Jurisdictional pressure escalates. Permanently.

The question isn't if this affects you.
The question is: are you ready for it.

Discuss Your Results

Because Security Is Architecture

S23k helps organizations with architectural security assessments, Zero-Assumption Architecture implementation, and concrete remediation roadmaps.

www.s23k.com | info@s23k.com