This guide draws on the Vitaver Playbook’s emphasis on verifiable results. Ask whether an authorized team can rebuild the service from the material it holds.
List everything the service needs to come back
A backup plan should let the recovery team rebuild a usable service when the production platform cannot be trusted. Start with one application and ask an engineer to list what they would need on a clean host. For a case management system, that might include a database and transaction logs, attachments, a compatible release of the application, configuration and infrastructure definitions, and a supported path to the operating system and dependencies. Identify licenses, recovery keys and credentials by owner and retrieval procedure. Store those secrets through a separately approved mechanism.
The CISA StopRansomware Guide recommends offline, encrypted copies of critical data, regular testing, and keeping system images and applicable software recovery material. Apply that approach to your application inventory. Preserve a tested pairing of application release and data schema. If you rely on a service provider, verify what its export captures and what you would need to import it somewhere you control. Ask a real operator to restore a representative export, including attachments and relationships the business needs.
Give the copies different jobs
- Online recovery copy. Use it for routine restores and short recovery windows. Maintain version history, restrict write and deletion privileges, and alert on changes to protection settings. Set retention using the history you need, the available storage, and the approved rules for that data.
- Protected off-site copy. Place a copy beyond the likely reach of the same physical event. Where a network-accessible vault uses retention locks, verify who can change the policy, who can read the backup, and whether production administrators can affect either. Geographic separation and administrative separation address different risks.
- Offline copy. Maintain an encrypted recovery set disconnected from routine production access after transfer. Give a designated custodian responsibility for rotation, condition checks, access logs, and replacement media. Keep a second authorized means of obtaining necessary decryption keys and document how the backup can be retrieved if the primary location or team is unavailable.
NIST’s contingency planning guide connects backup and alternate site strategies to recovery objectives. Choose capture frequency separately from retention. Capture frequency determines the best possible age of the latest usable copy. Retention gives investigators time to find an earlier copy if damage went unnoticed. A long retention period is useful only if the organization can locate, read, and trust the relevant versions. Replication can help you switch sites rapidly during an availability incident. Keep independent historical copies for events in which bad changes or deletions may also have replicated.
Rebuild without the production environment
Set up a recovery exercise in a controlled environment. Have an authorized operator retrieve a recovery set, verify it against an independently protected manifest, build the compatible runtime, and restore the selected data. Treat unexpected changes as evidence to investigate. For suspected compromise, involve the incident lead in selecting a restore point and checking the source images and restored environment before reconnecting anything to normal networks.
A database that starts is an intermediate checkpoint. Run tests for records and attachments, permissions, and a sample transaction. Check that the application works with the selected database schema and can be operated with the recovery credentials. Log the age of the recovered data at the simulated incident time and how long it took until a service owner could verify the result. Include the time spent obtaining keys, tools, and approvals.
Make the inventory maintainable
Assign a service owner to the recovery set and a separate reviewer to verify it. Record copy location, dependency versions, retention, last successful restore, and approved custodians. When a database schema, encryption key, license, or build process changes, review the affected entries and rerun the check before declaring the new version recoverable. If a proprietary component cannot be restored with the material you hold, assign someone to negotiate an export, obtain supported installers, or scope another recovery path. The acceptance check is a verified service rebuild from the documented set.