Skip to content

COMPARISON · Ways of working

Safety engineering vs security engineering

Organisations that keep safety and security as fully separate disciplines usually discover the gap between them during an incident, not before.

LAST REVIEWED

What is different

  • Adversary model: safety engineering assumes no intelligent adversary — failures are random, systematic or human error; security engineering assumes an adversary who adapts to defences.
  • Primary analysis method: safety uses hazard and risk analysis (e.g. HAZOP, FMEA, fault tree analysis); security uses threat modelling and attack surface analysis.
  • Governing standards: safety draws on IEC 61508, ISO 26262, IEC 62304 and sector equivalents; security draws on IEC 62443, ISO/SAE 21434, ISO/IEC 27001 and, for products, the CRA's essential requirements.
  • Failure mode: a safety failure typically degrades gracefully towards a defined safe state; a security failure can be silent, persistent and designed to avoid detection.
  • Organisational home: safety functions often sit with systems or functional safety engineering, close to certification bodies; security functions often sit with IT, product security or software engineering.

What overlaps

Both disciplines exist to prevent unacceptable outcomes and both rely on the same underlying discipline: define the system boundary, identify what can go wrong, assign a requirement to control it, implement it, and produce verification evidence that a reviewer or assessor can follow without reconstruction. Increasingly the outcomes converge directly — a cybersecurity attack on a safety-related control function is a safety hazard with a security cause, and standards such as IEC 61508's newer editions and IEC 62443 explicitly acknowledge this.

Both disciplines also produce a case-based argument for a regulator or customer: a safety case and a security case share a similar structure — claim, argument, evidence — even where the underlying analysis differs.

Which applies to you

  • Your product has no network connectivity and no software-controlled safety function: safety engineering practice alone may be sufficient, though this is increasingly rare.
  • Your product is purely IT and carries no risk of physical harm: security engineering practice, aligned to ISO/IEC 27001 or sector security standards, is the primary discipline.
  • Your product has both a safety function and a digital attack surface — most industrial, automotive, medical and defence products today: you need both disciplines working from a shared risk register, with an explicit interface where a security event can trigger a safety hazard.
  • You are unsure whether a given failure mode is a safety issue, a security issue, or both: treat it as both until analysis proves otherwise — the cost of under-scoping is higher than the cost of one extra review.

Frequently asked questions

Can the same engineer do both safety and security analysis?

Individuals can be competent in both, and cross-training is valuable, but the two disciplines require different mindsets and often different qualifications for formal sign-off. Most mature organisations keep distinct competence requirements while running a shared process for issues that span both.

Does IEC 61508 cover cybersecurity?

IEC 61508 addresses functional safety and, in its current edition, acknowledges that security threats can compromise safety functions, pointing towards security standards such as IEC 62443 for the security analysis itself. It does not replace a dedicated security standard.

Which comes first, the safety case or the security case?

Neither strictly precedes the other; they should be developed in parallel from the same system definition, with an explicit point where security risk assessment results feed into the safety hazard analysis for any safety function with a digital attack surface.

Is a security vulnerability always a safety hazard?

No. Many vulnerabilities affect confidentiality or availability without creating a path to physical harm. The determination requires tracing whether the vulnerable component has any influence, direct or indirect, over a safety-related function.