Safety engineers and security engineers often analyse the same industrial system in parallel and produce requirements that contradict each other, so a control that satisfies one discipline quietly defeats the other. Co-engineering means running functional safety and cyber security as one coordinated lifecycle: shared system models, linked hazard and threat analyses, agreed rules for resolving conflicts and a single change process. This guide explains how IEC 61508, IEC 61511 and the IEC 62443 series fit together (the subject of IEC TR 63069), how the same logic applies to ISO 26262 and ISO/SAE 21434 in vehicles, how to use STPA and STPA-Sec as a joint method, and how to measure whether the two disciplines are actually working together. It is for OT security leads, functional safety engineers, system integrators and product teams building safety-related systems.
Key takeaways
- A cyber attack is a hazard cause. IEC 61508 and IEC 61511 both expect malevolent action to be considered in hazard and risk analysis, and the UK HSE says conventional HAZOP is not sufficient on its own for cyber threats.
- Same system, two analyses, one model. Build one architecture and control structure model, then derive hazards and threats from it so every finding traces to the same components.
- Conflicts are predictable. Lockouts, patching, encryption latency and fail states produce the same disputes on every project; agree resolution rules before design starts.
- Security change is safety change. A patch, firewall rule or account change on a safety-related system needs a safety impact assessment, and a safety modification needs a security one.
- STPA gives a common language. Losses, hazards, unsafe control actions and loss scenarios work for both disciplines, which is why STPA-Sec is the most common bridge method.
Why safety and security must be engineered together
Functional safety asks whether a system can reach a dangerous state through random hardware failure, systematic error or misuse. Security asks whether someone can deliberately put it there. The physical consequence is the same, so the analyses overlap, but the assumptions differ: safety analysis traditionally treats failures as independent and probabilistic, while an attacker can cause several "independent" failures at once and adapts to the defences.
- Common cause by design: an attacker who compromises the engineering workstation can change both the basic process control system and the safety instrumented system. HSE operational guidance OG86 notes that standard hazard studies such as HAZOP generally do not address multiple simultaneous contingencies or the capabilities of malicious actors.
- Safety functions as targets: MITRE ATT&CK for ICS catalogues impacts such as T0880 Loss of Safety and T0837 Loss of Protection, and inhibit response techniques such as T0838 Modify Alarm Settings and T0816 Device Restart/Shutdown.
- Security controls as hazards: an account lockout on an operator HMI during an upset, a reboot to apply a patch, or an intrusion prevention system dropping a trip command can each cause harm.
- Shared evidence: both disciplines need an accurate asset inventory, architecture drawings, data flows and change records. Building them twice wastes effort and guarantees inconsistency.
For the SIS specific controls (separation from the BPCS, key switches, bypasses) see the Safety Instrumented Systems Security Guide; this page covers the engineering process that ties the disciplines together.
The standards landscape and how the pieces connect
| Standard | Discipline and scope | Integrity scale | Co-engineering hook |
|---|---|---|---|
| IEC 61508 (ed 2.0, 2010) | Functional safety of E/E/PE safety-related systems; basic safety publication | SIL 1 to 4 | Hazard and risk analysis is expected to include a security threats analysis where malevolent action is reasonably foreseeable |
| IEC 61511 | Safety instrumented systems in the process industry | SIL 1 to 4 | Clause 8.2.4 requires a security risk assessment of the SIS; HSE OG86 says following its process may meet that clause |
| IEC 62443 series | Security of industrial automation and control systems: asset owner programme (2-1), risk assessment for system design (3-2), system requirements (3-3), product lifecycle (4-1), component requirements (4-2) | Security levels SL 1 to 4 | Zones and conduits give the boundary that safety analysis can reuse |
| IEC TR 63069:2019 | Framework for functional safety and security in industrial-process measurement, control and automation (Technical Report, TC 65) | None of its own | Guidance on applying IEC 61508 and IEC 62443 together |
| ISO 26262 (2018) and ISO/SAE 21434 (2021) | Road vehicle functional safety and cybersecurity engineering | ASIL A to D; cybersecurity assurance levels | ISO describes them as complementary and often implemented together; both are marked to be revised |
| HSE OG86 | UK interpretive guidance for COMAH sites and NIS operators of essential services | Zone consequence and criticality | Ties IACS cyber risk directly to major accident hazards |
| NIST SP 800-82 Rev. 3 | Guide to OT security (September 2023) | None | OT security tailored to performance, reliability and safety requirements |
Medical devices follow the same pattern under IEC 62304 and IEC 81001-5-1; see the IEC 81001-5-1 Health Software Security Lifecycle Guide. For vehicles see the Automotive Cybersecurity Guide.
Aligning the two lifecycles
Both IEC 61508 and IEC 62443 are lifecycle standards. Co-engineering works when each phase has a defined hand-off artefact that the other discipline must review.
| Lifecycle phase | Safety activity | Security activity | Joint artefact or gate |
|---|---|---|---|
| Concept and scope | Define equipment under control, losses and hazards | Define system under consideration, assets and consequences | One system boundary and loss list |
| Hazard and threat analysis | HAZOP, LOPA, hazard and risk analysis | IEC 62443-3-2 high level and detailed risk assessment; zones and conduits | Cross-reference: each hazard lists credible cyber causes; each zone lists the safety functions it hosts |
| Requirements | Safety requirements specification with SIL per function | Target security level per zone and conduit; IEC 62443-3-3 system requirements | Conflict register signed by both leads |
| Design and realisation | Architecture constraints, diagnostics, proven in use | IEC 62443-4-1 secure development, hardening, component requirements (4-2) | Joint design review; security controls checked for timing and fail state |
| Verification and validation | Functional testing, proof tests, safety validation | Security testing, penetration testing on a test bed, not on a live SIS | Combined factory and site acceptance tests |
| Operation and maintenance | Proof testing, bypass management, management of change | Monitoring, patching, account management, incident response | Single management of change process with both impact assessments |
| Modification and decommissioning | Impact analysis and re-verification | Re-assessment of zone risk; credential and key revocation | Updated cross-reference and evidence pack |
STPA and STPA-Sec as the shared method
System-Theoretic Process Analysis (STPA) models a system as a hierarchy of feedback control loops and asks how control can become unsafe. The STPA Handbook (Leveson and Thomas, MIT) states it can be used for any emergent system property, including cyber security; STPA-Sec is the security-focused extension proposed by Young and Leveson.
- Define the purpose: losses (injury, environmental damage, loss of mission, loss of sensitive information) and system-level hazards that lead to them. Losses must not name components or causes.
- Model the control structure: controllers (operators, BPCS, SIS, remote services), control actions and feedback. It is a functional model and does not assume controllers obey.
- Identify unsafe control actions: for each control action, the four types are: not providing it causes a hazard; providing it causes a hazard; providing it too early, too late or out of order; stopping it too soon or applying it too long.
- Identify loss scenarios: why an unsafe control action could happen. Safety teams list incorrect feedback, flawed requirements and component failures; security teams add spoofed feedback, manipulated set points (ATT&CK for ICS T0831 Manipulation of Control) and unauthorised mode changes (T0858 Change Operating Mode).
Why it helps: both disciplines review the same unsafe control actions, so a scenario such as "trip command not provided when pressure exceeds limit" receives safety causes and attack causes side by side, and a single constraint covers both. It complements rather than replaces HAZOP, LOPA and IEC 62443-3-2.
Resolving conflicts between safety and security requirements
| Typical conflict | Safety position | Security position | Common resolution |
|---|---|---|---|
| Account lockout and session timeout on HMIs | Operators must never be locked out during an upset | Repeated failures and idle sessions should lock | No lockout on control room HMIs, physical access control to the room, shared view-only sessions, alerting on failures instead of locking |
| Patching | Any change may need re-verification | Known vulnerabilities must be fixed | Risk-based patching per zone, test bed validation, compensating conduit controls where patching waits for a turnaround |
| Encryption and authentication on field protocols | Added latency may breach the safety response time | Unauthenticated commands can be forged | Measure timing on the test bed; protect the conduit at the boundary where in-line authentication would break timing |
| Fail state of security devices | Fail safe: a failing firewall must not block a trip | Fail secure: a failing device must not open the network | Keep safety trip paths independent of security devices; decide fail state per conduit and record it |
| Physical security of exits | Doors must open for evacuation | Doors should stay locked | Fail-safe release on fire alarm with monitoring and response |
| Remote access for vendors | Fast specialist support during faults | Remote paths are a top entry route | Brokered, recorded, time-limited sessions approved by the site, read-only by default |
Decision guidance
- Precedence rule: where a security control could prevent a safety function from acting, safety wins and the security risk is treated by another layer. Where a missing security control lets an attacker defeat the safety function, the security control is itself safety-related.
- Small sites: one engineer may hold both roles; use a single combined risk register and an external reviewer for independence.
- Large operators: name a co-engineering authority (often the functional safety manager with the OT security lead) who signs the conflict register and the management of change.
- Product suppliers: publish the security assumptions the safety case depends on (for example "engineering port must be on a dedicated zone") so integrators inherit them.
Implementing co-engineering: a phased approach
- Inventory and boundary (weeks 1 to 6): agree one asset register and architecture, list every safety function, and map it to IEC 62443 zones and conduits. HSE OG86 treats a zone that could compromise a critical zone as critical itself, so consequence inherits until a real segregation point.
- Cross-reference analyses (months 2 to 4): add cyber causes to existing hazard records and safety functions to each zone's risk assessment. OG86 does not credit existing countermeasures when setting a zone's risk profile, which avoids circular reasoning.
- Pilot STPA-Sec on one unit: choose a unit with a high SIL function and remote connectivity; run a joint workshop with operations, safety and security.
- Merge change management: one management of change form with mandatory safety impact and security impact fields (NIST SP 800-53 CM-3 and CM-4 describe the security side).
- Test on a representative bed: verify that security controls do not change response times or fail states before site deployment; never test intrusive tools on a live safety system.
- Roll back: every security change on a safety-related zone has a documented reversal and a defined safe state if reversal fails, consistent with SI-17 Fail-safe Procedures and SC-24 Fail in Known State.
- Operate and review: include cyber scenarios in emergency drills and review the cross-reference at each turnaround or major modification.
Where identity fits
Engineering workstation logins, vendor accounts and the service accounts used by historians and remote access brokers are the paths through which attackers reach safety-related zones. Shared and default accounts on controllers are both a safety and a security concern because they remove accountability for changes. See the Identity for OT Environments Guide.
Detection, measurement and framework mapping
- Monitor: controller mode changes and program downloads, safety bypass and override events, unexpected trips, engineering workstation logins, new conduits or remote sessions, and alarm setting changes.
- Metric, coverage: share of safety functions with a documented cyber cause analysis and a zone assignment (target 100%).
- Metric, change discipline: share of changes on safety-related zones with both impact assessments signed before implementation.
- Metric, conflicts: open entries in the conflict register older than 90 days.
- Metric, investigation: share of unexplained trips or spurious shutdowns reviewed for cyber cause.
- Metric, remote access: vendor sessions into safety-related zones that were brokered and recorded.
| Framework | Reference | Co-engineering relevance |
|---|---|---|
| NIST SP 800-53 Rev. 5 | SA-8 Security and Privacy Engineering Principles; SA-11 Developer Testing and Evaluation | Engineering security into the lifecycle and testing it |
| NIST SP 800-53 Rev. 5 | RA-3 Risk Assessment; CM-3 Configuration Change Control; CM-4 Impact Analyses | Joint risk assessment and change control |
| NIST SP 800-53 Rev. 5 | SI-17 Fail-safe Procedures; SC-24 Fail in Known State; CP-12 Safe Mode | Defining what happens when controls fail |
| NIST CSF 2.0 | ID.RA-04, ID.RA-07, PR.PS-01, GV.RM-03 | Impact identification, change risk, configuration management, enterprise risk integration |
| MITRE ATT&CK for ICS | T0880 Loss of Safety, T0837 Loss of Protection, T0838 Modify Alarm Settings, T0858 Change Operating Mode | Attack outcomes the cross-referenced analysis must cover |
| IEC 62443 | 2-1, 3-2, 3-3 | See the IEC 62443-3-2 OT Risk Assessment Guide |
Common failure patterns
- Separate registers: the hazard log and the security risk register describe the same system with different asset names, so nobody can show which threats affect which safety functions.
- Security controls that break safety timing: deep packet inspection or encryption added to a trip path after validation, without re-testing response time.
- Assuming the SIS is independent because it is a different box: shared engineering workstations, shared networks or shared vendor accounts reintroduce common cause.
- Patching outside management of change: a well-meant update on a safety-related host that was never impact assessed by the safety team.
- Ignoring IT dependencies of operations: plants can be stopped without any OT compromise when business systems the operation depends on fail, as covered below.
Lessons from incidents
In the Colonial Pipeline ransomware attack, attackers entered through a dormant VPN account without MFA, and the company shut down its pipeline as a precaution from 7 to 13 May 2021 once ransomware was found; public reporting does not describe attackers reaching pipeline control systems. The co-engineering lesson is that a credible joint analysis must model dependencies between business IT and operations, and pre-agree which security events justify a precautionary shutdown, so the decision is made from analysis rather than uncertainty. The 2017 TRITON attack on a refinery safety system, covered in the Safety Instrumented Systems Security Guide, shows the direct case: attackers targeting the safety layer itself.
Practitioner checklist
- Keep one asset register and architecture model used by both safety and security teams.
- Add credible cyber causes to every hazard record for safety-related functions.
- Map every safety function to an IEC 62443 zone and record the zone's target security level.
- Run a joint STPA-Sec workshop on at least one high integrity unit.
- Maintain a signed conflict register with a precedence rule for safety and security.
- Require safety and security impact assessments on every change to safety-related zones.
- Validate security controls for response time and fail state on a test bed before deployment.
- Prohibit account lockout on control room HMIs and compensate with physical and monitoring controls.
- Broker and record all vendor remote access to safety-related zones.
- Include cyber scenarios and precautionary shutdown criteria in emergency exercises.
Standards and references
- IEC: TR 63069 Framework for functional safety and security
- IEC: 61508-1 Functional safety of E/E/PE safety-related systems, General requirements
- ISA: ISA/IEC 62443 Series of Standards
- HSE: OG86 Cyber Security for Industrial Automation and Control Systems
- MIT PSASS: STPA Handbook
- NIST: SP 800-82 Rev. 3 Guide to Operational Technology Security
- ISO: ISO/SAE 21434 Road vehicles, Cybersecurity engineering
- ISO: ISO 26262-1 Road vehicles, Functional safety
Related NHI Mgmt Group resources: Safety Instrumented Systems Security Guide · IEC 62443 Guide · IEC 62443-3-2 OT Risk Assessment Guide · Identity for OT Environments Guide