Cyber Recovery Exercises: Testing Recovery in a Ransomware Scenario

Does the recovery plan work in a real scenario? An exercise assuming AD and the backup console are lost: scenario options, business-service-based scope, stage timings and recording findings.
Technical summary
Technology
Backup software, Active Directory, isolated recovery network, NIST SP 800-84
Topic
A technical recovery test with a ransomware scenario
Level
Intermediate–advanced
Prerequisites
  • A dependency list of critical business services
  • An immutable or offline backup copy
  • Approved RTO and RPO targets
In this article7 sections

Passing a routine restore test doesn’t mean recovery after a ransomware attack is ready. In a real incident, the attacker has usually become domain admin, got into the backup console and encrypted the hypervisors; recovery is done without trusting any part of this environment. This guide explains how to set up the scenario, scope, success criteria and measurement of a technical recovery exercise that assumes an attack.

The difference between a restore test and a cyber recovery exercise

A classic restore test assumes the infrastructure is intact. The backup server is up, the console is entered with a domain account, the restored machine is connected to the production network; DNS, DHCP and authentication are in place. This test is part of the “0 errors” goal in the 3-2-1-1-0 rule and is explained in the 3-2-1-1-0 backup rule and restore testing article. Necessary but not enough.

A cyber recovery exercise asks other questions. If the backup server has also been encrypted, where do we rebuild the console from, and in how many hours? Where’s the backup encryption password, and who can access it? If Active Directory has been compromised, which identity will we use to get into restored servers? Who will decide which recovery point is clean, and based on what? None of these questions come up in a routine test.

Dimension Routine restore test Cyber recovery exercise
Backup console Assumed to be working Lost or untrustworthy
Identity infrastructure Production AD is used AD is assumed compromised
Network Production network Isolated recovery network
Recovery points The latest backup A point verified as clean
What’s measured Readability of the data Recovery time of the business service

Scenario assumptions

The value of the exercise from an enterprise cybersecurity standpoint depends on how ruthlessly the assumptions are written. An exercise with soft assumptions only gives confidence, not information. At least the following should be written explicitly in the scenario document:

  • Identity: All domain admin accounts and the KRBTGT key have been compromised; no recovery system is signed in to with existing AD accounts.
  • Backup infrastructure: The backup server and configuration database can’t be used; only the immutable repository, the offline copy or the copy in object storage is accessible.
  • Virtualization: Production hypervisors and shared storage have been encrypted; recovery is done on separate hardware or in a separate cluster.
  • Timeline: The attacker’s initial access happened some time before encryption; backups from the last few days may contain a persistence mechanism.
  • Access: Corporate email and the on-premises password vault can’t be used; team communication and critical information are provided through an out-of-band channel.

It isn’t necessary to apply all assumptions at once in the first round. Depending on the team’s maturity, start with one of the variations below:

Variation Main assumption Capability tested
A: Console loss No backup server Rebuilding the console and importing the repository
B: Identity loss AD compromised Recovering a DC in the isolated environment, resetting accounts
C: Dirty latest backup Malware in the latest backups Choosing a recovery point and scanning
D: Full scenario A + B + C and loss of hypervisors End-to-end recovery of a critical business service

In practice, the most efficient order is to start with A and add one variation each round; that way each bottleneck shows up separately. In the field, even variation A stops most teams in the first hour: the installation media, license file or repository access details aren’t at hand. For the scanning and decision method in the dirty latest backup variation, see the choosing a clean recovery point guide.

Scope and success criteria

Define the scope by business service, not by a server list. Writing “ERP system” isn’t enough; a verifiable sentence such as “users can sign in to the ERP, create orders and issue invoices” is written. Then the service’s dependency chain is mapped: domain controller, DNS, database server, application server, license server, file share, certificate services. An exercise that starts without a dependency map typically locks up in the same place: the application server starts, can’t find the license server, and the service doesn’t come up. That’s the stopping point we often see.

For each scope item, these criteria are written before the exercise:

  • The target recovery time (RTO) and acceptable data loss (RPO), approved by the business unit.
  • The checks to be done to prove the recovery point is clean.
  • The functional test the application owner will run for the service to count as “working”.
  • The conditions for stopping the exercise; for example, a suspected isolation breach.

Exercise steps and role assignment

The technical flow should be the same as the recovery flow in a real incident. This skeleton works in most mid-sized environments:

  1. Kickoff and communication: The exercise lead explains the scenario, and the timestamped record starts. The team switches to the out-of-band communication channel.
  2. Preparing the isolated environment: The recovery network, a temporary hypervisor and a management workstation are set up, or a pre-prepared environment is brought up. For design principles, see the isolated recovery environment design article.
  3. Rebuilding the backup console: The backup server is installed from clean installation media, and the immutable repository or object storage is imported. For encrypted backups, the password is taken from escrow.
  4. Recovery point decision: A candidate point is chosen based on the timeline, scanned, and the decision is recorded.
  5. Recovering the identity infrastructure: A domain controller is restored in the isolated environment and the forest recovery steps are applied. Details are in the AD forest recovery after ransomware guide.
  6. Recovering the application layer: Database, application and file servers are brought up in dependency order.
  7. Functional verification: The application owner runs the pre-written test.
  8. Close: The environment is cleaned up, records are collected and a findings meeting is planned.

Role assignment should include at least these people: an exercise lead who runs the scenario and keeps time; practitioners responsible for backup, virtualization and identity infrastructure; application owners; and a recorder who writes down decisions. An observer notes steps the practitioners skip and doesn’t intervene. In small teams, one person takes on more than one role; that’s unavoidable. But don’t make the practitioner and the recorder the same person. Someone wrestling with the console doesn’t write down times.

Durations and observations to measure

A single “total recovery time” figure doesn’t tell you what to fix. Keep times stage by stage so the bottleneck shows:

Measurement Start End
Environment preparation time The scenario starts The isolated network is usable
Console recovery time Installation starts Backup chains are visible
Decision time Candidate points are listed The recovery point is approved
Identity recovery time The DC restore starts Accounts can sign in
Service recovery time The scenario starts The functional test passes

Next to the times, write qualitative observations too: what information couldn’t be found, which password wasn’t in escrow, which document was out of date, at which step the team didn’t know what to do. The findings that come up again and again in exercises are well known: the backup encryption password being recorded only on the backup server; the backup software license file being inaccessible; local administrator passwords of restored servers being unknown; the database restore taking much longer than expected because of repository performance; the application depending on hard-coded IP addresses.

For the measurement to be reliable, the recorder should write the start and end of each stage against the same time source. The backup software’s task logs and hypervisor event records are a second source of evidence to verify verbal accounts. Memory misleads; logs don’t.

Turning findings into improvements

The output of the exercise is a findings record. Each finding is tied to a remediation task with an owner, a date and a verification method. This template can be used to keep records consistent:

Finding No   : EX-2026-04
Stage        : Console recovery
Observation  : The backup encryption password is recorded only in the
               configuration on the old backup server; not in escrow.
Impact       : Encrypted backup chains couldn't be imported;
               the stage was delayed by 2 h 40 min.
Root cause   : No password escrow procedure is defined.
Action       : Passwords will be put in sealed escrow and a separate vault.
Owner        : Backup administrator
Target date  : 2026-12-15
Verification : In the next exercise, the import will be done using
               the password taken from escrow.

Divide the findings into three groups: technical (requiring architectural change), process (documentation such as runbooks and escrow) and competence (the team’s training needs). Technical findings often require budget; when you go to management with measured times, the decision comes quickly. An example calculation: target RTO 8 hours, measured service recovery time 19 hours. The stage durations in the table show in which stage the gap arose, and investment is directed to that stage.

The next exercise is set up to verify that previous findings have really been closed. An unclosed finding eats up the same hours again in the next incident. To integrate the exercise program with the incident response plan, see our cyber incident response readiness and exercises service, and to close gaps in the backup architecture, the backup security and ransomware readiness page. For the planning and evaluation method, NIST’s SP 800-84 guide to test, training and exercise programs is a solid framework.

Frequently asked questions

Does a cyber recovery exercise put the production environment at risk?

Not if it’s set up correctly. Restored machines carry the same names, IP addresses and domain as production; that’s why restores are done on an isolated network with no routing, VPN or management connection to production. The backup repository is accessed for reading only, and when the exercise ends, all machines in the test environment are deleted. The real risk is forgetting the isolation rule; verify the network separation separately before the exercise.

How often should the exercise be done?

At least once a year for a full-scope scenario; and additionally when there’s a major change in the backup architecture, identity infrastructure or critical applications. Narrow-scope parts, for example bringing up only a domain controller in the isolated environment, can be repeated every three months.

If routine restore tests succeed, is a cyber exercise still needed?

Yes. A routine test shows whether a single server comes back while the backup console, Active Directory and network services are intact. In a cyber scenario, those dependencies themselves are gone; rebuilding the console, accessing encryption passwords and bringing up the identity infrastructure from scratch can only be tested in an exercise.

What criterion should be used to consider the exercise successful?

Not a single server starting, but a pre-written business service working end to end. For example, a criterion such as “users can sign in to the ERP and an order record can be created” is defined together with a target time and a data loss limit; the actual values are compared with these targets.

Support

Contact us for support or a discovery session on this topic.

Contact form
[email protected]

Eretek Cybersecurity

Firewall, VPN, endpoint security and threat prevention. We reduce the attack surface and measure and report the risks.

Knowledge Center

Other articles