​

MOD Security Compliance

• /arckit.mod-secure - MOD Secure by Design assessment (JSP 440, continuous assurance)
• /arckit.jsp-936 - JSP 936 AI Assurance Documentation (defence AI systems)

​

Command: /arckit.mod-secure

​

What is MOD Secure by Design?

​

Usage

/arckit.mod-secure "Logistics Management System OFFICIAL"
/arckit.mod-secure 001  # By project ID

• Project name or ID

​

Output: ARC-{PROJECT_ID}-SECD-MOD-v1.0.md

• 7 MOD Secure by Design Principles assessment
• NIST Cybersecurity Framework (CSF) assessment (5 functions)
• Three Lines of Defence evaluation
• Classification-specific requirements (OFFICIAL → TOP SECRET)
• CAAT (Cyber Activity and Assurance Tracker) self-assessment
• Supplier security attestation requirements

​

Critical Changes (Post-August 2023)

• Cyber security is a licence to operate - cannot be traded out
• Applies to ALL new programmes and systems
• Legacy systems transition when accreditation expires (completed by 31 March 2024)
• Supplier-owned continuous assurance (not MOD accreditation)
• Suppliers must attest that systems are secure (ISN 2023/10)
• Senior Responsible Owners (SROs), capability owners, and delivery teams are accountable

​

The 7 MOD Secure by Design Principles

​

Principle 1: Understand and Define Context

• Understand the capability’s overall context
• How it will use and manage MOD data
• How it achieves its primary business/operational outcome

• Context documented (mission, users, data flows)
• Data classification determined
• Operational environment understood
• Stakeholder security requirements captured

​

Principle 2: Apply Security from the Start

• Security embedded in design from inception (not bolt-on)
• Security requirements defined early
• Security architecture designed before build

• Security requirements in initial specifications
• Threat model created in Discovery/Alpha
• Security architecture reviewed and approved
• Security expertise involved from start

​

Principle 3: Apply Defence in Depth

• Multiple layers of security controls
• Fail-safe defaults (secure by default)
• Assume breach (design for compromise)

• Layered security controls (network, host, application, data)
• Segmentation and least privilege implemented
• Monitoring and detection at each layer
• Containment and recovery capabilities

​

Principle 4: Follow Secure Design Patterns

• Use proven secure architectures
• Leverage NCSC/NIST guidance
• Avoid known insecure patterns

• NCSC Secure Design Principles applied
• NIST CSF controls mapped
• Common vulnerabilities (OWASP Top 10) mitigated
• Secure coding standards followed

​

Principle 5: Continuously Manage Risk

• Risk assessment is ongoing (not one-time)
• Risk register maintained through-life
• Security testing continuous

• Risk register actively maintained
• Regular vulnerability scanning and pen testing
• Security incidents tracked and lessons learned
• Continuous monitoring and threat intelligence

​

Principle 6: Secure the Supply Chain

• Third-party components assessed
• Vendor security requirements in contracts
• Software supply chain protected

• Software Bill of Materials (SBOM) maintained
• Third-party risk assessments completed
• Supplier security attestations obtained (ISN 2023/10)
• Open source software vetted
• Supply chain attack vectors mitigated

​

Principle 7: Enable Through-Life Assurance

• Security posture maintained post-deployment
• Regular security reviews
• Capability to respond to new threats

• Security monitoring operational
• Incident response capability proven
• Patching and update process defined
• Security governance continues through-life
• Decommissioning process includes secure data deletion

​

NIST Cybersecurity Framework Assessment

​

Identify

• Asset inventory (hardware, software, data, people)
• Business environment and criticality
• Governance structure and policies
• Risk assessment methodology

​

Protect

• Access control (authentication, authorisation)
• Data security (encryption at rest/in transit, DLP)
• Protective technology (firewalls, AV, IDS/IPS)
• Security awareness training

​

Detect

• Continuous monitoring (SIEM, SOC integration)
• Anomaly and event detection
• Security testing (vulnerability scanning, pen testing)
• Detection processes and procedures

​

Respond

• Incident response plan
• Communications and reporting (to MOD CERT)
• Analysis and mitigation
• Improvements from lessons learned

​

Recover

• Recovery planning (backup, DR, BC)
• Improvements (post-incident review)
• Communications and restoration

​

Three Lines of Defence

• Delivery Team Security Lead (DTSL) appointed
• Security requirements owned by capability owner
• Day-to-day security management

• Technical Coherence Assurance
• Security policies and standards
• Independent security reviews

• Internal audit of security controls
• Penetration testing by independent teams
• External audit (NAO, GIAA)

​

CAAT (Cyber Activity and Assurance Tracker)

1. Register on CAAT
   • Every programme must register in Discovery/Alpha
   • CAAT is the self-assessment tool for cyber security maturity
   • Available through MOD Secure by Design portal (DefenceGateway account required)
2. Appoint Delivery Team Security Lead (DTSL)
   • DTSL owns security for delivery team (First Line of Defence)
   • May also appoint Security Assurance Coordinator (SAC)
   • Project Security Officer (PSyO) still required for SECRET+ systems
3. Complete CAAT self-assessment question sets
   • Based on the 7 MOD Secure by Design Principles
   • Assess cyber security maturity throughout lifecycle
   • Regular updates required (not one-time submission)
4. Complete Business Impact Assessment (BIA)
   • Understand criticality and impact of compromise
   • Informs risk assessment and security controls
5. Implement security controls
   • Based on NIST CSF, NCSC guidance, JSP 440 requirements
   • Defence in depth approach
   • Continuous improvement throughout lifecycle
6. Conduct continuous security testing
   • Vulnerability scanning (regular, automated)
   • Penetration testing (at key milestones)
   • Security audits by Third Line of Defence
7. Maintain continuous risk management
   • Risk register actively maintained
   • Threats and vulnerabilities continuously monitored
   • Security incidents tracked and lessons learned applied
8. Supplier attestation (for systems delivered by suppliers)
   • Suppliers must attest that systems are secure (ISN 2023/10)
   • Supplier-owned continuous assurance (not MOD accreditation)
   • Supplier security requirements in contracts
9. Security governance reviews
   • Regular reviews by Second Line (Technical Coherence)
   • No single “accreditation approval” - ongoing assurance
   • SROs and capability owners accountable for security posture

​

Classification-Specific Requirements

​

OFFICIAL

• Cyber Essentials baseline
• Basic access controls and encryption
• Standard MOD security policies

​

OFFICIAL-SENSITIVE

• Cyber Essentials Plus
• MFA required
• Enhanced logging and monitoring
• DPIA if processing personal data

​

SECRET

• Security Cleared (SC) personnel minimum
• CESG-approved cryptography
• Air-gapped or assured network connectivity
• Enhanced physical security
• CAAT assessment and security governance review before deployment

​

TOP SECRET

• Developed Vetting (DV) personnel
• Compartmented security
• Strictly controlled access
• Enhanced OPSEC measures

​

Report Contents

1. Executive Summary
   • Overall security posture
   • 7 principles compliance score
   • NIST CSF maturity
   • Critical issues
   • CAAT self-assessment status
2. 7 MOD SbD Principles Assessment
   • Each principle assessed: ✅/⚠️/❌
   • Evidence gathered
   • Gaps identified
   • Recommendations
3. NIST CSF Assessment
   • 5 functions assessed
   • Maturity level for each function
   • Gaps and recommendations
4. Three Lines of Defence
   • First Line: DTSL appointed, security owned
   • Second Line: Assurance activities
   • Third Line: Independent audit status
5. Classification Requirements
   • Data classification determined
   • Classification-specific controls assessed
   • Personnel security clearances verified
6. CAAT Self-Assessment
   • Registration status
   • Question sets completed
   • Maturity level
   • Actions required
7. Supplier Security Attestation
   • Supplier attestations obtained (ISN 2023/10)
   • Supplier security requirements in contracts
   • Third-party risk assessments completed
8. Action Plan
   • Critical priority (0-30 days) - deployment blockers
   • High priority (1-3 months) - significant risk reduction
   • Medium priority (3-6 months) - continuous improvement

​

Key MOD Standards

• JSP 440: Defence Manual of Security (primary security policy)
• JSP 440 Leaflet 5C: Secure by Design mandate (August 2023)
• JSP 453: Digital Policies and Standards for Defence
• ISN 2023/09: Industry Security Notice - Secure by Design Requirements
• ISN 2023/10: Industry Security Notice - Supplier attestation and legacy accreditation withdrawal
• IAMM: Information Assurance Maturity Model (8 domains, 0-5 scale)

​

Resources

• MOD Secure by Design
• MOD Secure by Design Portal - Requires DefenceGateway account
• CAAT - Self-assessment tool (DefenceGateway access)
• JSP 440
• JSP 453
• ISN 2023/09 - Secure by Design Requirements
• ISN 2023/10 - Supplier attestation
• NCSC Secure Design Principles
• Defence Digital

​

Command: /arckit.jsp-936

​

What is JSP 936?

​

Usage

/arckit.jsp-936 "Target Recognition AI OFFICIAL-SENSITIVE"
/arckit.jsp-936 001  # By project ID

• System name or project ID

​

Output: ARC-{PROJECT_ID}-J936-v1.0.md

​

JSP 936 Framework

• Human-Centricity
• Responsibility
• Understanding
• Bias & Harm Mitigation
• Reliability

• Critical (20-25 points)
• Severe (15-19 points)
• Major (10-14 points)
• Moderate (5-9 points)
• Minor (1-4 points)

• Planning
• Requirements
• Architecture
• Algorithm Design
• Model Development
• Verification & Validation
• Integration & Use
• Quality Assurance

• Critical/Severe: Ministerial (2PUS) approval
• Major: Defence-Level (JROC/IAC) approval
• Moderate/Minor: TLB-Level (delegated) approval

​

AI Ethical Risk Classification

• Human safety and wellbeing
• Operational effectiveness
• Legal and ethical compliance
• Public trust and reputation
• International obligations

• Technology maturity (TRL)
• Data quality and availability
• Algorithm complexity
• Operational environment
• Human factors and training

​

The 5 Ethical Principles (Detailed)

​

Principle 1: Human-Centricity

• Assess and consider impact on humans
• Ensure positive effects outweigh negatives
• Human-AI interaction design
• Stakeholder engagement
• Cognitive load considerations
• Trust calibration

​

Principle 2: Responsibility

• Ensure meaningful human control
• Clear accountability
• Human-in-loop/on-loop/out-of-loop decision framework
• Decision authority matrix
• Incident response ownership

​

Principle 3: Understanding

• Relevant personnel understand how AI functions
• Explainability requirements (SHAP, LIME, Grad-CAM)
• Training programme (AI literacy, system-specific)
• Documentation (model cards, operating procedures)
• Performance boundaries documented

​

Principle 4: Bias and Harm Mitigation

• Proactively identify and reduce unintended biases
• Training data representativeness
• Performance disparities across groups
• Fairness metrics
• Harm identification (direct, indirect, systemic)
• Mitigation strategies

​

Principle 5: Reliability

• Demonstrate robust, secure performance
• Performance bounds defined (design domain)
• Robustness testing (adversarial resilience)
• Failure modes and effects analysis (FMEA)
• Security (AI-specific threats: adversarial examples, data poisoning, model theft)

​

The 8 AI Lifecycle Phases (Detailed)

​

Phase 1: Planning

• AI use case justification
• Algorithm development roadmap
• Data strategy
• Resource plan
• Stakeholder map
• Initial ethical risk assessment
• Governance structure (RAISO, Ethics Manager)

​

Phase 2: Requirements

• Functional requirements
• Non-functional requirements (performance, safety, security)
• Ethical requirements (derived from 5 principles)
• Safety requirements (from hazard analysis)
• Security requirements (AI-specific threats)
• Acceptance criteria
• Hazard analysis (HAZOP, FMEA)

​

Phase 3: Architecture

• System architecture
• AI pipeline architecture (data → preprocessing → model → postprocessing)
• Deployment architecture
• Traceability matrix (requirements → components)
• Failure modes and graceful degradation
• Security architecture (threat model)
• Human-AI interface design

​

Phase 4: Algorithm Design

• Algorithm selection and justification
• Design decisions (architecture, hyperparameters, trade-offs)
• Verification methods
• Output verification (sanity checks)
• Edge case handling
• Explainability design

​

Phase 5: Model Development

• Training data (sources, size, characteristics, provenance)
• Training process (procedure, hyperparameters)
• Model card (performance, limitations, intended use)
• Performance evaluation (metrics on train/val/test sets)
• Bias analysis (performance across subgroups)
• Uncertainty calibration
• Reuse considerations

​

Phase 6: Verification & Validation (V&V)

• Test plan
• Verification testing (does it meet specs?)
• Validation testing (does it meet user needs?)
• Edge case testing
• Adversarial testing
• User acceptance testing
• V&V report (results, pass/fail, issues)

​

Phase 7: Integration & Use

• Integration testing
• Operational acceptance testing
• Deployment procedures
• Operational documentation
• User training delivered
• Monitoring and alerting setup
• Incident response procedures

​

Phase 8: Quality Assurance

• Continuous monitoring
• Performance tracking
• Model drift detection
• Retraining schedule
• Audit trail
• Continuous improvement

​

Governance Structure

• TLB-level appointed
• Quarterly review of AI portfolio
• Accountability for AI systems in TLB

• Embedded in project team
• Day-to-day ethics oversight
• Escalates issues to RAISO

• Annual review by external ethics board
• Technical assurance by Defence Science & Technology Laboratory (Dstl)

​

Report Contents

1. Executive Summary
   • AI components identified
   • Ethical risk classification (Critical → Minor)
   • Approval pathway required
   • Overall compliance with 5 principles
   • Go/No-Go decision
2. AI Component Discovery
   • All AI/ML systems identified and documented
   • Purpose, input, output, human involvement
   • Training data, model type, deployment context
3. Ethical Risk Classification
   • Likelihood × Impact assessment
   • Risk score and classification
   • Approval pathway determination
   • Unacceptable risk criteria check
4. 5 Ethical Principles Assessment
   • Each principle assessed in detail
   • Evidence gathered
   • Gaps identified
   • Recommendations
5. 8 Lifecycle Phases Documentation
   • Each phase comprehensively documented
   • Assurance activities completed
   • Traceability maintained
   • Phase gate readiness
6. Governance and Oversight
   • RAISO appointed and engaged
   • Ethics Manager assigned
   • Independent assurance plan
   • Approval pathway documented
7. Action Plan
   • Critical actions (0-30 days)
   • High priority (1-3 months)
   • Continuous improvement (3-6 months)

​

Integration with Other MOD Commands

• /arckit.mod-secure - MOD security requirements feed into JSP 936 Principle 5 (Reliability, security section)
• /arckit.ai-playbook - Civilian AI Playbook overlaps with JSP 936 ethical principles
• /arckit.atrs - ATRS publication for MOD AI systems (if not classified SECRET+)

​

Resources

• JSP 936 Part 1: Directive
• JSP 936 Part 2: Guidance
• Defence AI Strategy
• Dstl AI Assurance

​

Example Use Cases

/arckit.jsp-936 "Autonomous Target Recognition System"  # Comprehensive lifecycle documentation

/arckit.jsp-936 001  # Update documentation, re-assess risks, continuous improvement

/arckit.jsp-936 001  # Generate evidence pack for independent ethics review