Cyber resilience starts from an assumption: the incident will happen
Cybersecurity seeks to prevent the incident. Cyber resilience prepares the organisation to survive it. The two complement each other, but they involve neither the same people nor the same evidence. The first is judged by what does not happen. The second is judged by what you know how to do on the day everything stops.
Yet many continuity arrangements disappoint when they are needed. The plans exist, often voluminous, sometimes old. They describe an infrastructure that has changed, mention people who left long ago and assume backups that nobody has restored in full. The weakness is not in the drafting. It is in the absence of testing.
The move to the cloud adds a misunderstanding. Many teams take it for granted that the high availability promised by the provider amounts to a recovery plan. It covers the failure of a piece of equipment or a site. It covers neither a handling error, nor the encryption of data by an attacker, nor the failure of the provider itself.
Continuity, recovery, crisis: three plans that do not replace one another
The business continuity plan answers a business question: how do you continue to serve customers if the usual tool is unavailable? It provides for degraded modes, manual procedures and priorities.
The disaster recovery plan answers a technical question: in what order and by what means is the information system brought back into service? It relies on two objectives set by the business: the maximum tolerable interruption time and the maximum tolerable data loss.
The crisis management arrangements answer a decision question: who decides, with what information, and who speaks to customers, authorities and employees?
These three plans rest on prior work, the business impact analysis, which ranks activities according to what their interruption costs. The ISO 22301 standard, devoted to business continuity management systems, provides a reference framework for the whole.
In a ransomware attack scenario, two points become decisive. Backups must be out of reach of production administration accounts, otherwise they are encrypted along with everything else. And recovery requires a clean environment: the directory and administration tools are rebuilt first, before the applications.
DORA and NIS2: what the texts require
Two European texts have made digital resilience an obligation for many organisations. They differ in both nature and scope.
| DORA | NIS2 | |
|---|---|---|
| Reference | Regulation (EU) 2022/2554 | Directive (EU) 2022/2555 |
| Nature | Regulation, directly applicable | Directive, to be transposed by each Member State |
| Scope | Financial entities listed in Article 2, and oversight framework for critical third-party ICT service providers | Essential and important entities in the sectors listed in Annexes I and II |
| Application | Applicable since 17 January 2025 | According to national law |
| Main areas | ICT risk management, management, classification and reporting of incidents, digital operational resilience testing, risks related to third-party ICT service providers, information sharing | Governance (Article 20), risk management measures (Article 21), reporting obligations in the event of a significant incident (Article 23) |
Three common features directly concern operational management.
- The responsibility of senior management. Article 5 of DORA places responsibility for the ICT risk management framework on the management body. Article 20 of NIS2 provides that management bodies approve the cybersecurity risk management measures and oversee their implementation.
- Continuity as an expected measure. Article 21 of NIS2 cites business continuity (for example backup management and disaster recovery), crisis management and supply chain security. Chapter II of DORA deals in particular with response and recovery, backup and restoration.
- Proof through testing. Chapter IV of DORA is devoted to digital operational resilience testing.
The two texts fit together: Article 1(2) of DORA specifies that, for financial entities identified as essential or important entities under NIS2, the regulation is considered a sector-specific Union legal act within the meaning of Article 4 of that directive. Notification deadlines and thresholds are set by the texts and their implementing acts. We do not summarise them here. Our entries on the DORA regulation and the NIS2 directive present their structure.
Third parties, a dependency to be mapped
A critical activity rarely depends on your teams alone. It relies on a cloud provider, a software-as-a-service vendor, a managed service provider, and on each one’s own subcontractors. Chapter V of DORA organises the management of this risk. Its Article 28 provides in particular for a register of information on contractual arrangements and exit strategies for services supporting critical or important functions.
Beyond the obligation, three questions deserve a written answer for each critical provider. What happens if it stops tomorrow? How many of your critical activities rely on it, directly or indirectly? With what, and under what conditions, could you replace it? When these questions intersect with those of location and applicable law, they connect with our page on sovereign and hybrid cloud.
Our approach: start from activities, prove through exercises
Our consultants work within your teams; decisions remain yours.
Having the business set the objectives. The impact analysis is carried out with the departments concerned. A tolerable interruption time chosen by IT alone commits nobody.
Tracing the chain of dependencies. For each critical activity, we trace the applications, data, infrastructure and third parties, down to single points of failure.
Measuring actual capability. A restoration test, even a partial one, gives an observed duration. That is what we compare with the objective, not the duration announced by the plan.
Closing the gap in order of criticality. Depending on the case, the answer is a standby architecture, an isolated backup, an automated rebuild using infrastructure code, a contractual clause or a business procedure.
Exercising, then doing it again. A crisis exercise involves senior management. Each exercise produces a report, actions and a date for the next one.
Building resilience into the life cycle. Every new application is given a continuity class from its design. Everyday incidents are handled according to the practices of our SRE and observability offering, whose post-incident reviews feed into the plans.
The most common blind spots
- The plan never run. A document approved in committee says nothing about the actual duration of a restoration.
- Replication mistaken for backup. Corrupted or encrypted data is replicated as faithfully as sound data.
- Crisis tools that depend on the disaster. If email and the directory are affected, the crisis unit must have independent means of communication and documents accessible offline.
- Software-as-a-service applications outside the scope. Their availability and the recovery of your data must appear in the plans in the same way as your own applications.
- Compliance treated as paperwork. Producing the required documents without changing actual capabilities leaves the risk intact.
What allows you to say an organisation is ready
Together with you, we choose observable indicators, with no promised value:
- the share of critical activities whose recovery objectives are validated by a business owner;
- the share of critical applications that have undergone a successful restoration test over the period;
- the gap between the target recovery time and the time measured during tests;
- the share of critical providers covered by an exit strategy and reviewed clauses;
- how long ago the last crisis exercise involving senior management took place;
- the share of actions arising from exercises closed by their due date.
The third indicator is the most useful to present to senior management: it replaces an assurance in principle with an observation.