From DORA requirements to engineering controls
The Digital Operational Resilience Act applies to EU financial entities and reaches deep into the ICT services that support them. Legal teams interpret its obligations; engineering teams have to make them true in systems, pipelines and runbooks. This article maps the main requirement areas to the technical controls that usually implement them.
Ce contenu est actuellement disponible uniquement en anglais.
Scope and roles
DORA applies to a wide range of EU financial entities – credit institutions, payment and e-money institutions, investment firms, insurers and others – and establishes an oversight framework for ICT third-party service providers designated as critical. Engineering providers that are not themselves financial entities are affected through contracts: their clients must ensure that outsourced ICT services meet DORA’s requirements, so they flow down into service agreements, audit rights and technical expectations.
This article does not interpret the law. It assumes that your compliance team has decided which obligations apply, and shows the engineering controls that commonly provide evidence for them.
ICT risk management: know what you run
Every risk-management framework starts with an inventory. DORA and its risk-management standard expect entities to identify and classify ICT assets, their dependencies and the business functions they support. In engineering terms this means:
- A service catalogue generated from infrastructure-as-code and runtime discovery, not from a spreadsheet updated once a year.
- Dependency maps derived from traces and network flows, showing which services support which critical or important functions.
- Data classification attached to stores and topics, so that controls can be applied by sensitivity.
When the inventory is generated automatically, it stays current and can answer an auditor’s question – “which systems support this function, and who changed them last month?” – with evidence rather than recollection.
Protection and change control
The risk-management standard covers ICT security policies, access control, encryption and cryptographic key management, network security, vulnerability and patch management, and ICT change management. Most of these map to controls that a modern delivery pipeline can enforce:
- Changes to production happen only through version-controlled pipelines with review, automated tests and recorded approvals.
- Access is granted through identity providers with multi-factor authentication, least privilege and time-limited elevation for emergencies.
- Dependencies and images are scanned continuously, with defined remediation times by severity.
- Keys and secrets live in a managed vault or HSM, with rotation and access logs.
The point is not to add bureaucracy but to make the existing engineering process produce evidence as a side effect.
Detection, logging and incident classification
DORA requires entities to detect anomalous activity, manage ICT-related incidents and report major ones to authorities within defined timeframes. Delegated Regulation (EU) 2024/1772 specifies criteria for classifying incidents – such as the number of clients affected, duration, geographical spread, data losses, criticality of services affected and economic impact.
Engineering teams can make classification fast and reliable by capturing the inputs automatically: which services and functions an alert concerns, how many clients or transactions are affected, when impact began and ended. Incident tooling with structured fields for these criteria – filled in from monitoring data where possible – turns a stressful judgement call into a checklist. Centralised, tamper-resistant logs with defined retention are the foundation for both detection and post-incident analysis.
Resilience testing
DORA requires a digital operational resilience testing programme that includes, as appropriate, vulnerability assessments, scenario-based tests, performance and end-to-end testing, and – for entities identified by authorities – threat-led penetration testing. For engineering teams, the practical programme looks familiar:
- Load and performance tests against defined objectives before significant releases.
- Failover and restore exercises with measured recovery times and data loss.
- Dependency-failure and chaos experiments on critical paths.
- Regular vulnerability scanning and penetration tests with tracked remediation.
Each test should produce a dated report with findings and remediation owners. The programme’s value lies in the follow-up, not in the test itself.
Third-party risk and exit strategies
Entities must manage risks from ICT third-party providers, including cloud platforms and software vendors, and maintain exit strategies for services supporting critical or important functions. Engineering choices determine whether an exit is realistic. Infrastructure as code, portable container images, standard protocols and data exports in documented formats make migration possible; deep use of proprietary services without abstraction makes it slow and risky.
An exit plan worth having is tested at least partially: restore a backup into a different environment, deploy the service on alternative infrastructure, or run a dependency in a degraded mode for a defined period.
Evidence as a product
Auditors and supervisors ask for evidence. Teams that treat evidence as a by-product of good engineering – pipeline logs, test reports, incident timelines, access reviews generated from systems of record – spend far less time on audits than teams that reconstruct it manually. A small internal “evidence catalogue” mapping each requirement area to the system that produces its evidence is a useful artefact to build early.
Business continuity and backups
DORA also expects ICT business continuity policies, response and recovery plans, and backup and restoration procedures with defined recovery objectives. In engineering terms these are the familiar building blocks of resilience: recovery point and recovery time objectives agreed per critical function, replication and backup designs that meet them, and restore procedures that are tested on a schedule rather than assumed.
Two details deserve attention. First, backups must be protected from the incidents they are meant to recover from – ransomware, compromised credentials, faulty automation – which usually means immutable or logically separated copies with their own access controls. Second, restoring must be possible in a reasonable time: for large data sets, restore duration can exceed the recovery objective on its own, which is a finding to address in design, not during an outage.
Recording each restore test with its date, scope, measured duration and outcome produces exactly the kind of evidence that reviews ask for.
Exemple: Making incident classification fast
A payment institution’s incident process relied on a duty manager estimating impact by hand. The engineering team added structured fields to the incident tool – affected services, mapped critical functions, start time, client and transaction counts – populated automatically from alert labels and traffic metrics. A classification helper applied the criteria defined by the compliance team and suggested whether the incident might be major. In the next significant outage, the compliance officer had the data needed for the initial assessment within minutes instead of hours, and the post-incident report was assembled from the tool’s timeline.
Pièges courants
- Treating DORA as a documentation exercise detached from real systems.
- Asset inventories maintained by hand and therefore always outdated.
- Resilience tests without tracked remediation.
- Incident data that must be reconstructed after the fact.
- Exit strategies that have never been tried in any form.
Questions et réponses
Does DORA apply to software vendors?
DORA applies directly to financial entities and establishes oversight for ICT third-party providers designated as critical. Other providers are affected through contractual requirements set by their financial-sector clients. Your legal team should assess your specific position.
Is chaos engineering required?
DORA requires a testing programme proportionate to the entity; it does not prescribe specific tools. Controlled failure experiments are one effective way to test scenarios and recovery.
Where should we start?
With the inventory and dependency map of critical functions. Almost every other requirement – testing, incident classification, third-party risk – depends on knowing what supports what.
Liste de contrôle
- Generate the ICT asset and dependency inventory automatically.
- Map services to critical or important functions.
- Enforce change control through pipelines with recorded approvals.
- Centralise logs with defined retention and integrity protection.
- Capture incident classification inputs automatically.
- Run a resilience testing programme with tracked remediation.
- Design for and partially test exit from key providers.
Sources primaires
- Regulation (EU) 2022/2554 on digital operational resilience for the financial sector (DORA)
- Commission Delegated Regulation (EU) 2024/1774 – ICT risk management tools, methods, processes and policies
- Commission Delegated Regulation (EU) 2024/1772 – classification of ICT-related incidents and cyber threats
- European Supervisory Authorities (EBA, EIOPA, ESMA) – DORA regulatory and implementing technical standards