Most disaster recovery plans assume the disaster is an accident, so they fail when the disaster is an adversary who has already taken the identity systems, backup consoles and replication links the plan depends on. Disaster recovery (DR) was built to restore service after a fire, flood, hardware failure or regional outage, and it measures success in recovery time and recovery point. Cyber recovery is the capability to restore service after a deliberate, destructive attack, when you cannot trust the production environment, the most recent data or the administrators' accounts. The two share tools but rest on opposite assumptions about trust. This guide explains where those assumptions diverge, why DR runbooks break under ransomware, what a cyber recovery capability adds (clean rooms, identity-first rebuild order, realistic recovery times and a defined hand-off with incident response) and how to split ownership and budget between the two. It is for CISOs, infrastructure and resilience leads, and incident response managers.
Key takeaways
- The difference is the trust assumption. DR assumes the systems and people running the recovery are sound; cyber recovery assumes they may be compromised until proven otherwise.
- Replication copies the attack. Synchronous and near-synchronous replication, the backbone of fast DR, carries encrypted or corrupted data and attacker persistence to the secondary site within seconds.
- Identity is rebuilt first, not last. NIST SP 800-184 notes that an identity or authentication server frequently has to be recovered before files, and an attacker who keeps directory access defeats the recovery.
- Recovery time is set by investigation, not infrastructure. Restoring before persistence is understood risks restoring the attacker; realistic cyber recovery times are counted in days to weeks for critical services.
- Run both, owned separately but tested together. DR stays with infrastructure and continuity; cyber recovery needs joint ownership with security and incident response, and its own budget line.
The short answer
Disaster recovery restores the latest good state as fast as possible. Cyber recovery restores a verified clean state, into an environment the attacker cannot reach, in an order that cuts the attacker off. If your DR plan can bring back a data centre in four hours but uses the same domain, the same backup administrators and the same replication stream as production, it is a DR plan only. It will probably not survive ransomware with domain administrator access.
For the full continuity picture, including business impact analysis and plan types, see the Business Continuity and Disaster Recovery Guide. This page covers only the decision between the two recovery models and how they fit together.
Side-by-side comparison
| Dimension | Disaster recovery | Cyber recovery |
|---|---|---|
| Threat model | Accidental or natural loss: site, hardware, power, region, human error | Deliberate destruction by an adversary with privileged access: encryption, wiping, backup deletion |
| Trust assumption | Production identities, tools and recent data are trustworthy | Nothing connected to production is trusted until verified |
| Recovery point | The latest copy, ideally seconds old | The latest copy verified clean, which may be days or weeks old |
| Recovery target | Secondary site or region, often warm and replicated | An isolated recovery environment or clean room with separate identity |
| Data protection | Replication, snapshots, standard backups | Immutable, logically or physically isolated copies with integrity checks |
| Identity | Shared with production; DR site often joined to the same directory | Separate administrative identities and a rebuilt or recovered directory |
| Trigger | Infrastructure outage declared by operations | Security incident declared by incident response, with recovery gated by investigation |
| Speed driver | Infrastructure capacity and failover automation | Forensic confidence, integrity scanning and identity rebuild |
| Evidence handling | Not a concern | Must not destroy forensic evidence or alert the attacker |
| Typical owner | Infrastructure and business continuity | Shared between security, incident response and infrastructure |
Why DR plans fail against ransomware
NIST SP 800-184, the Guide for Cybersecurity Event Recovery, warns that starting recovery too early may alert the adversary that the intrusion has been found, and that without a good understanding of the adversary's objective and high confidence in its persistence mechanisms, recovery has a high chance of being ineffective or inefficient. Most DR plans were written without that constraint. The common breaking points are:
- Replication carries the damage: the secondary site receives encrypted files, corrupted databases and attacker tools as fast as production does. Failover brings up the same compromised state somewhere else.
- The DR site shares an identity plane: if the secondary site authenticates against the same directory, an attacker with domain administrator rights there owns both sites. See Recovering IAM and PAM After Ransomware or Domain Compromise.
- Backups are reachable and deletable: CISA's #StopRansomware Guide notes that many ransomware variants attempt to find and delete or encrypt accessible backups. A backup console joined to the domain and managed with domain credentials is part of the attack surface. MITRE ATT&CK tracks this as T1490 Inhibit System Recovery, alongside T1486 Data Encrypted for Impact and T1485 Data Destruction.
- The runbook depends on what is down: DR runbooks often assume single sign-on, the password vault, the ticketing system, internal DNS and email are working. In a destructive attack they may be the first casualties.
- Recovery points are too recent: DR tooling selects the latest restore point by default. If the attacker was present for weeks, the latest point contains their persistence.
- No hand-off with incident response: operations teams restore systems that responders still need for evidence, or bring services back before credentials have been reset.
For the attacker behaviour itself, see the T1490 Inhibit System Recovery guide.
What cyber recovery adds
Cyber recovery is not a product category. It is a set of capabilities that DR plans usually lack, built on the DR estate where possible.
| Capability | What it means in practice | Reference |
|---|---|---|
| Isolated, immutable copies | Copies that no production identity can delete or alter, with retention set in advance and deletion delayed and alerted | NCSC principles for ransomware-resistant cloud backups (Principle 1); Cyber Recovery Vault Guide |
| Out-of-band access | A way into the backup system that does not depend on production identity, so one account cannot lock everyone out | NCSC Principle 2 |
| Versioned restore points | Backups kept by time period rather than count, so a flood of corrupted copies cannot push out the good ones | NCSC Principle 3 |
| Integrity verification | Scanning and validating restore points before use | NIST CSF 2.0 RC.RP-03 |
| Clean room | An isolated environment with its own identities and tooling where restore points are tested and services rebuilt | Cloud Recovery and Clean Room Guide |
| Rebuild assets | Golden images, infrastructure as code templates and installation media kept offline | CISA #StopRansomware Guide |
| Identity-first sequence | A documented order that restores a trusted directory and administrative path before applications | NIST SP 800-184 |
| Change alerting | Alerts on mass deletions, stopped backups and retention or encryption changes, delivered even if production is compromised | NCSC Principle 5 |
Clean rooms: where recovery actually happens
A clean room is the answer to the question "restore to where?" Restoring into the compromised production network risks re-infection; CISA advises adding only verified clean systems to a new recovery network segment and identifying any dropper malware before rebuilding from backups. A workable clean room has:
- Separate identity: administrative accounts and authentication that do not trust the production directory.
- Its own tooling: scanning, forensic and deployment tools that were not installed from, or managed by, production.
- Controlled connectivity: no route to production until a service is declared clean; outbound access limited to what recovery needs.
- Enough capacity for the minimum viable service set: not the whole estate, but the services the business cannot survive without. See the Minimum Viable Company and Critical Service Recovery Order Guide.
- Exit criteria: what must be true for a rebuilt service to move from the clean room to production use.
The DR site can host the clean room if it can be cut off from production identity and networks on demand. Many organisations find it cannot, which is the main architectural reason cyber recovery needs its own design work.
Identity rebuild order
NIST SP 800-184 states that it is frequently necessary to recover an identity or authentication server before recovering files, and that services such as mail may depend on directory services. It also notes that if an adversary keeps access to a directory or management account, recovery can fail. A DR plan usually restores identity as one more server; cyber recovery treats it as the foundation.
- Recovery foothold: clean administrative workstations and break-glass accounts that never depended on production.
- Directory: restore or rebuild the directory from a restore point predating the compromise, in isolation, and verify it before anything trusts it. See the Identity Recovery Guide.
- Privileged access: restore the vault and privileged access tooling, then rotate privileged credentials.
- Core services: DNS, PKI, time and key management, which most applications assume.
- Machine identities: service accounts, API keys, certificates and cloud access keys used by the services about to come back.
- Applications and data: in the order set by the critical service list.
- User credentials: CISA advises issuing password resets for affected systems only after the environment has been fully cleaned and rebuilt, so the new credentials are not captured.
RTO realism
A DR recovery time objective is an engineering figure: how long failover and restore take. In a cyber event, the clock is set by things the DR plan does not measure:
- Scoping: how long it takes to establish when the attacker arrived and which restore points predate them.
- Integrity checking: scanning restore points and images before use (NIST CSF 2.0 RC.RP-03).
- Identity rebuild: recovering and verifying the directory and rotating privileged and machine credentials.
- Restore throughput: isolated and immutable copies are often on slower storage than replicated DR copies.
- Deadlines: NIST SP 800-184 notes that ransomware or extortion deadlines may force recovery to proceed on incomplete information, which is a business decision that needs an owner.
Set a separate cyber recovery time objective for each critical service, based on rehearsal results rather than DR figures, and report both to the board. A four-hour DR objective and a ten-day cyber recovery objective for the same service are not a contradiction; they describe different events.
Incident response hand-offs
NIST SP 800-184 says initial restoration should be tactical, to avoid alerting the adversary or destroying forensic evidence. NIST CSF 2.0 expresses the same link in RC.RP-01: the recovery portion of the incident response plan is executed once initiated from the incident response process. In practice that needs explicit hand-offs.
| Hand-off | From | To | Gate |
|---|---|---|---|
| Declare cyber recovery, not DR | Incident commander | Recovery lead | Evidence of malicious destruction or privileged compromise |
| Preserve before restore | Recovery lead | Forensics | Images and logs of affected systems captured before rebuild |
| Select restore point | Forensics | Recovery lead | Earliest compromise date and persistence mechanisms identified with stated confidence |
| Release service to production | Recovery lead | Service owner | Integrity verified, credentials rotated, monitoring in place (RC.RP-05) |
| Declare recovery complete | Incident commander | Executive sponsor | Agreed criteria met and documentation completed (RC.RP-06) |
For the end-to-end response sequence, see the Ransomware Response and Recovery Guide.
Ownership and budget split
- Disaster recovery: owned by infrastructure or IT operations under the business continuity programme. Budget covers secondary capacity, replication, failover automation and DR testing.
- Cyber recovery: owned jointly by the CISO and the head of infrastructure, with the incident response lead as the operational authority during an event. Budget covers isolated and immutable copies, the clean room, separate administrative identities, integrity scanning, offline rebuild assets and full rehearsals.
- Shared: the critical service list, recovery order, dependency maps and the joint exercise programme. One list, maintained once.
Avoid funding cyber recovery by cutting DR. Accidental outages remain more frequent, and DR is still the right answer for them. Where the budget is tight, reuse DR capacity for the clean room only if it can be isolated from production identity on demand, and spend first on isolated copies and identity separation, which DR cannot provide.
Decision guidance
- Small organisation, mostly SaaS: DR is largely the providers' responsibility. Cyber recovery focuses on independent, immutable backups of SaaS data and configuration, separate admin accounts and a documented identity recovery path.
- Mid-size, hybrid estate: keep existing DR; add an immutable copy outside the production identity domain, a small clean room sized for the critical service set and a rehearsed identity rebuild.
- Large or regulated enterprise: run a formal cyber recovery programme with its own objectives, a dedicated vault and clean room, and joint DR and cyber exercises at least annually. Map it to sector expectations.
- Heavy cloud: treat the cloud control plane as part of the threat model; see the Cloud Recovery and Clean Room Guide.
Implementation: a phased approach
- Assess: run a tabletop that starts from "an attacker holds domain administrator and backup administrator rights" and walk the DR plan step by step. Record every step that depends on a compromised system.
- Separate identity for recovery: move backup administration and recovery tooling to accounts outside the production directory, with phishing-resistant MFA.
- Protect copies: add immutable, isolated copies for the critical service set, with deletion delays and alerts.
- Build the clean room: start small, with capacity for identity and two or three critical services; prove it can be isolated.
- Write the cyber runbook: identity-first order, IR hand-offs and exit criteria, kept offline.
- Rehearse: recover a critical service from an immutable copy into the clean room, time each stage and fix the gaps. Roll back by keeping the existing DR plan unchanged until the cyber runbook has passed a rehearsal.
- Extend and repeat: add services in order of criticality and rehearse at least annually.
The Disaster Recovery Testing Guide covers how to run the exercises.
Detection and measurement
- Backup tamper alerts: mass deletion, retention change, stopped jobs and encryption setting changes, routed to a channel outside production.
- Time to clean restore point: how long to identify and verify the most recent clean copy of a critical service in a rehearsal.
- Time to trusted identity: hours from declaration to a verified directory and privileged access path in the clean room.
- Cyber recovery time achieved vs objective: per critical service, from the latest rehearsal.
- Coverage: share of critical services with isolated, immutable copies and a tested cyber runbook.
- Timely restorations: percentage of timely restorations from backup, one of the quality-of-recovery metrics NIST SP 800-184 suggests.
Common failure patterns
- Calling DR failover a ransomware plan: the secondary site receives the same encrypted data and compromised directory.
- Backup consoles on the production domain: the attacker deletes recovery copies with the same credentials used to encrypt production.
- Restoring before scoping: services come back with the attacker's persistence and are lost again, or the attacker is alerted and escalates.
- Resetting passwords too early: new credentials are captured on systems not yet cleaned.
- Forgetting machine identities: service accounts and keys restored from backup still carry credentials the attacker holds.
- Rehearsing only the DR case: a test that assumes working single sign-on, vault and DNS proves nothing about a destructive attack.
Framework mapping
| Framework | Reference | Disaster recovery | Cyber recovery |
|---|---|---|---|
| NIST CSF 2.0 | PR.DS-11 | Backups created and tested | Backups protected against privileged deletion |
| NIST CSF 2.0 | RC.RP-01, RC.RP-03, RC.RP-05, RC.RP-06 | Partly applies | Recovery initiated from IR, integrity verified, end declared on criteria |
| NIST CSF 2.0 | RS.MI-01, RS.MI-02 | Not applicable | Containment and eradication gate recovery |
| NIST SP 800-53 Rev. 5 | CP-2, CP-4, CP-6, CP-7 | Contingency plan, testing, alternate storage and processing sites | Same controls, tested against a destructive scenario |
| NIST SP 800-53 Rev. 5 | CP-9, CP-10 | System backup, recovery and reconstitution | Protected backups and reconstitution to a known secure state |
| NIST SP 800-53 Rev. 5 | IR-4, IR-8 | Limited | Incident handling and plan that include recovery |
| MITRE ATT&CK v20 | T1490, T1486, T1485 | Not modelled | Attacker actions the design must survive |
Practitioner checklist
- Run a tabletop that assumes domain and backup administrator compromise against the current DR plan.
- Move backup administration out of the production directory, with phishing-resistant MFA.
- Keep immutable, isolated copies of the critical service set, with deletion delays and out-of-band alerts.
- Stand up a clean room with separate identity, sized for the minimum viable service set.
- Document an identity-first recovery order covering directory, privileged access, core services and machine identities.
- Agree IR hand-offs and gates, including evidence preservation before rebuild.
- Set and report separate cyber recovery time objectives per critical service.
- Keep golden images, infrastructure as code templates and runbooks offline.
- Rehearse a full recovery into the clean room at least annually and time each stage.
Standards and references
- NIST: SP 800-184 Guide for Cybersecurity Event Recovery
- NIST: SP 800-34 Rev. 1 Contingency Planning Guide for Federal Information Systems
- NIST: Cybersecurity Framework 2.0
- NIST: SP 800-53 Rev. 5
- CISA: #StopRansomware Guide
- NCSC: Principles for ransomware-resistant cloud backups
- MITRE ATT&CK: Inhibit System Recovery (T1490)
Related NHI Mgmt Group resources: Cyber Recovery Vault Guide · Business Continuity and Disaster Recovery Guide · Cloud Recovery and Clean Room Guide · Recovering IAM and PAM After Ransomware or Domain Compromise · Minimum Viable Company and Critical Service Recovery Order Guide