CAF Principle B5: Resilient Networks and Systems

CAF principle B5 asks whether you can keep essential functions running and restore them quickly after an attack or failure. For the Cyber Security and Resilience Bill (CSRB), resilience limits how severe an incident becomes, which decides whether it must be reported.

Last updated · Precursor Security
Where this stands

Status checked : the Bill passed the House of Lords on 2 November 2026 and returns to the Commons to consider Lords amendments (no date set). On 4 November 2026 the Commons extended its carry-over, so it lapses if not passed by the end of 29 April 2027. It is not yet law. References are to HL Bill 62 (as amended on Report), so wording and numbering may still change in the final Act. This guide is general information, not legal advice.

Key points

  • B5 has three outcomes: resilience preparation, design for resilience and backups.
  • Achieved requires continuity and recovery plans tested for practicality, plus physically separated critical systems.
  • Backups must stay accessible after an extreme event such as ransomware.
  • Under HL Bill 62, an attack stopped before it affected your systems is not a mandatory report, including for data centres.

What does CAF principle B5 cover?

Principle B5 asks whether you can keep your essential functions running, and restore them quickly, when systems are attacked or fail. It covers recovery planning, resilient design and backups. In the National Cyber Security Centre's Cyber Assessment Framework (CAF), it is the bridge between protecting systems and responding to incidents. For organisations preparing for the Cyber Security and Resilience Bill (CSRB), B5 directly affects how severe an incident becomes, which is what decides whether it must be reported.

The CAF is a framework, not a law, but NIS regulators use it to judge whether security duties are met.

What are the B5 contributing outcomes?

B5 has three contributing outcomes, each rated not achieved, partially achieved or achieved.

Outcome What it asks, in plain English
B5.a Resilience Preparation Do you know what you need to restore, in what order, and have you tested your continuity and recovery plans?
B5.b Design for Resilience Are critical systems segregated, cut off from internet services, and free of single points of failure?
B5.c Backups Are complete, secured, tested backups available even after an extreme event such as ransomware?

What is the difference between partially achieved and achieved?

B5.a Resilience Preparation. Partially achieved means you know every system and underlying technology needed to restore the essential function, understand how they depend on each other, and know the order in which to recover them. Achieved builds on that: your business continuity and disaster recovery plans have been tested for practicality, effectiveness and completeness using a mix of methods, such as manual failover, tabletop exercises or red teaming. You also use threat intelligence to put temporary extra measures in place when risk rises, for example during a widespread outbreak of destructive malware.

B5.b Design for Resilience. Partially achieved means critical systems are logically separated from business systems, for example in a separate network zone, cannot reach internet services such as web browsing and email, and resource limitations such as limited bandwidth or single network paths have been identified, if not yet fixed. Achieved means critical systems are separated by technical and physical means, such as separate infrastructure. All resource limitations are mitigated, geographical weaknesses are addressed, for example by replicating systems in another location and giving key connectivity alternative physical paths, and these assessments are kept up to date.

B5.c Backups. Partially achieved means backups are secured, cover data, configuration, software, equipment, processes and knowledge, would be accessible after an extreme event including ransomware, and are routinely tested. Achieved means backups are comprehensive, automatic and tested, held at central or secondary sites, and documented and reviewed routinely for everything needed to recover the essential function.

Where do organisations usually fall short?

Paraphrasing the CAF's not-achieved indicators:

  • No recovery map. A limited understanding of what is needed to restore critical systems.
  • Plans that have never been tested. Business continuity and disaster recovery plans incomplete, or never tested to see whether they work in practice.
  • Flat or connected networks. Critical systems not separated, or able to browse the web and use email.
  • Unmanaged single points of failure. Resource limitations, such as a single network path, not understood or planned for.
  • Backups that would not save you. Coverage is incomplete, backups are too infrequent, or restoring takes too long for the essential function.

What evidence could show B5 is achieved?

These are our suggestions, not NCSC requirements:

  • A recovery dependency map and restoration order for each essential function, with recovery time and recovery point objectives
  • Business continuity and disaster recovery plans with dated test reports from failovers, tabletop exercises and technical recovery tests
  • A "heightened threat" playbook listing temporary measures and who can trigger them
  • Architecture showing physical or infrastructure separation of critical systems, and evidence they cannot reach the internet
  • A single-point-of-failure register covering bandwidth, network paths, sites and suppliers, with mitigations
  • Backup coverage reports, immutable or offline copies, and timed full-restore tests from a ransomware scenario
  • Records of reviewing all of the above after significant system changes

A worked example

This is our illustration, not an NCSC case study. A regional airport's baggage and flight information systems run in a single on-site server room. Backups go nightly to a storage array in the same room, joined to the same domain. The disaster recovery plan has never been tested.

A ransomware attack that encrypts the domain could take out both live systems and backups, and the untested plan gives little confidence in recovery time. That is likely to be not achieved for B5.a and B5.c. To improve, the airport would replicate key systems to a second site, keep offline or immutable backup copies outside the main domain, and run a timed recovery exercise from a ransomware scenario.

How does B5 relate to the Cyber Security and Resilience Bill?

  • Resilience reduces impact, not the duty to report. Under the Bill, whether an incident must be reported would depend on whether it has actually affected the operation or security of your systems and whether its impact is, or is likely to be, significant. Factors include the extent and duration of disruption, users affected and geographical area (regulations 11, 12A and 14E). Good resilience shortens disruption, but if an incident has affected your systems and the impact is still significant, an initial notification would be due within 24 hours.
  • Near misses are not mandatory reports. HL Bill 62 removed the earlier wording that would have covered events merely "capable of" an adverse effect, and dropped mandatory near-miss reporting for data centres. An attack your resilient design stopped before it affected your systems is not a reportable incident under the current text. The Government has committed to a voluntary near-miss reporting mechanism, which does not yet exist. See breach vs incident vs near miss.
  • Security duties. Operators of essential services must take measures to prevent and minimise the impact of incidents under regulation 10 of the NIS Regulations. HL Bill 62 removes the words "with a view to ensuring the continuity of those services" from regulation 10(2), but resilience remains central to minimising impact. Relevant managed service providers would have a matching duty under new regulation 14B.
  • Customer updates. Under the Bill, data centres, RDSPs and RMSPs would have to notify likely affected UK customers after their full notification and then send updates as needed. Clear recovery estimates make those updates useful. See customer notification.

Where next?

Previous principle: B4 System Security. Next principle: B6 Staff Awareness and Training. Recovery planning continues in D1 Response and Recovery Planning. For the bigger picture, read CAF 4.0 explained, the CAF self-assessment guide and our overview of the framework.

Sources

Frequently asked questions

B5.a Resilience Preparation, B5.b Design for Resilience and B5.c Backups. Each can be rated not achieved, partially achieved or achieved. Together they cover knowing what to restore and in what order, designing systems to withstand failure and attack, and keeping backups you can rely on after an extreme event such as ransomware.

Partially achieved needs secured, routinely tested backups covering data, configuration, software and knowledge, accessible after an extreme event such as ransomware. Achieved needs comprehensive, automatic, tested backups at central or secondary sites, documented and routinely reviewed. Test full restores, not just backup jobs.

Not necessarily. Under the Bill, if an incident has affected the operation or security of your systems and its impact is, or is likely to be, significant, it would have to be reported. Resilience reduces disruption and duration, which are factors in that judgement, but it does not cancel the duty.

No, under the current text. HL Bill 62 removed mandatory near-miss reporting for data centre operators at Lords Report Stage. The Government has committed to a voluntary near-miss mechanism for all regulated organisations instead, which is still to be set up.

Get ready

Know where you stand before the duties land.

Take the free readiness assessment to see how your organisation measures up against the Cyber Security and Resilience Bill, or talk to our team about a plan.