Organisations that back up every night still fail CIS Control 11 when they cannot show which assets are in scope, that the backups are protected as well as the original data, that one copy is out of an attacker's reach, and that restores actually work. CIS Control 11 (Data Recovery) in CIS Controls v8.1 sets the minimum process, backup frequency, protection, isolation and testing an organisation should meet. This guide explains its five Safeguards by Implementation Group, the cadences and fail conditions in the CIS assessment specification, an implementation sequence, the evidence assessors expect and how the Control maps to NIST, ISO and ATT&CK. It is for infrastructure, backup and GRC teams. For backup architecture in depth (3-2-1-1-0, object lock options, vault design), see the Backup Security and Immutable Backup Guide; this page stays on what Control 11 requires and how to prove it.
Key takeaways
- Almost all of Control 11 is IG1. A data recovery process (11.1), automated backups (11.2), protection of recovery data (11.3) and an isolated instance of recovery data (11.4) apply to every organisation; only recovery testing (11.5) starts at IG2.
- The cadences are hard numbers. Backups run weekly or more often depending on data sensitivity, the process is reviewed at least annually, and the assessment specification fails 11.5 if more than three months pass between recovery tests.
- Backups must be as well protected as the source. Safeguard 11.3 asks for controls equivalent to the original data, typically encryption or data separation, which means the keys and the backup console need production-grade protection.
- Isolation is what survives ransomware. Attackers target recovery points before encrypting (ATT&CK T1490 Inhibit System Recovery), so 11.4's isolated instance needs separate credentials and immutability, not just a second copy on the same platform.
- Backup identities are tier 0 non-human identities. Backup service accounts and cloud keys can read every system and often delete every recovery point; their scope decides whether 11.3 and 11.4 hold.
What it is and who it applies to
CIS states the aim of Control 11 as establishing and maintaining data recovery practices sufficient to restore in-scope enterprise assets to a pre-incident and trusted state. Two words in that sentence drive the work: "in-scope", because the organisation must decide and document which assets and data are covered, and "trusted", because a restore of an infected or tampered backup is not a recovery.
- IG1: 11.1 to 11.4. A small organisation can meet IG1 with a written one-page recovery process, automated backups of its file shares, endpoints where needed and SaaS data, encryption switched on in the backup tool, and one copy held in an immutable cloud vault or offline media.
- IG2: adds 11.5, quarterly recovery testing on a sample of in-scope assets, with records.
- IG3: has no additional Control 11 Safeguards. At IG3 the differences are depth: more frequent backups for sensitive data, full-system and tier 0 restore tests, and tighter links with Control 17 (Incident Response Management).
CIS Controls v8.1 (2024) revised asset classes, added a Govern security function, realigned mappings to NIST CSF 2.0 and clarified some Safeguard descriptions. The Control 11 Safeguard numbers and titles match v8; the IG assignments, cadences and measures on this page come from the v8 CIS Controls Assessment Specification. Take asset class and security function values from the v8.1 download if you report them formally.
Structure and key requirements
| Safeguard | Title | IG | Minimum cadence | What it requires |
|---|---|---|---|---|
| 11.1 | Establish and Maintain a Data Recovery Process | IG1, IG2, IG3 | Review annually or on significant enterprise change | A documented process covering the scope of recovery activities, recovery prioritisation and the security of backup data |
| 11.2 | Perform Automated Backups | IG1, IG2, IG3 | Weekly or more often, based on data sensitivity | Automated backups of in-scope enterprise assets |
| 11.3 | Protect Recovery Data | IG1, IG2, IG3 | Continuous | Recovery data protected with controls equivalent to the original data, such as encryption or data separation |
| 11.4 | Establish and Maintain an Isolated Instance of Recovery Data | IG1, IG2, IG3 | Continuous | An isolated copy, for example version-controlled backup destinations in offline, cloud or off-site systems or services |
| 11.5 | Test Data Recovery | IG2, IG3 | Quarterly or more often | Recovery tests on a sample of in-scope enterprise assets |
What each Safeguard means in practice
- 11.1 three required elements: the assessment specification scores the process on whether it addresses scope, recovery prioritisation and backup security, and fails the Safeguard if there is no process or it is more than 12 months since the last update. Prioritisation means a recovery order driven by business impact analysis, with a recovery point objective (RPO, how much data loss is acceptable) and recovery time objective (RTO, how long the service can be down) per tier. NIST SP 800-34 Rev. 1 includes a business impact analysis template. See the Business Continuity and Disaster Recovery Guide.
- 11.2 weekly is the floor: the specification counts in-scope assets with authorised, properly configured backup software and those with a successful backup within seven days. Databases and file servers holding active business data usually need daily backups or continuous log shipping to meet a realistic RPO.
- 11.3 equivalent controls: the measure is the number of assets whose backup software encrypts backups. Equivalence also means access control: if only the finance team can read the ledger, a backup operator should not be able to browse restored copies of it freely.
- 11.4 isolated instance: the specification checks that backup destinations send at least one copy to an isolated instance. Isolation means an attacker who owns production identity cannot reach or delete it: offline media, a separate cloud account or tenant, or a vault with immutability that even administrators cannot override.
- 11.5 sampling and integrity: restore a sample of backups to a temporary location and check they work as expected. The specification measures backups tested, those that worked, those that did not, and the time since the last test (fail above three months).
Choosing an isolation approach for 11.4
| Approach | How isolation is achieved | Fits | Watch for |
|---|---|---|---|
| Offline or removable media (tape, rotated disks) | Physically disconnected | Small sites, tier 0 copies, long retention | Slow restores; media handling and storage; rotation discipline |
| Cloud vault with compliance-mode lock (for example AWS Backup Vault Lock in compliance mode, Azure Backup immutable vault "Enabled and locked") | WORM retention no user can shorten once locked | Most IG1 and IG2 organisations | Irreversible: wrong retention costs money; AWS compliance mode needs at least 72 hours of grace time before it becomes immutable |
| Governance-mode lock or Azure "Enabled" (unlocked) | Blocks deletion except by privileged users who can lift it | Pilot stage before locking | Not immutable: a stolen privileged identity can remove it |
| Separate backup account or tenant with its own identity provider | Different credentials and control plane | Larger cloud estates | Cross-account trust and copy roles must be one-way and append only |
| Same-platform snapshots | None from the platform's own administrators | Fast operational restores only | Does not meet 11.4 on its own |
The UK NCSC's principles for ransomware-resistant cloud backups give a useful test for whichever option you choose: backups resilient to destructive actions, no way for an attacker to deny all customer access, restore from an older version even if later versions are corrupted, robust key management, and alerts on significant changes or privileged actions. See the Cyber Recovery Vault Guide.
Identity and NHI requirements
Control 11 is where identity failures turn a ransomware incident into a total loss. Three areas need explicit treatment in the 11.1 process.
- Backup service accounts and agents: backup software reads every in-scope system, so its accounts often hold domain-wide or root-level rights. Scope them to read for backup and write to the target, deny delete and retention changes, vault their credentials, block interactive logon and alert on use from anywhere but the backup servers. See the Backup Identity Isolation Guide.
- Cloud keys with storage write access: in the Codefinger S3 ransomware campaign, attackers used leaked or stolen long-lived AWS access keys with object read and write permissions to re-encrypt bucket contents with server-side encryption using customer-provided keys (SSE-C), then set lifecycle rules to delete the files within seven days. Without an independent protected copy, recovery depended on the attacker's key. Replace long-lived keys with short-lived roles, block SSE-C where not needed, and keep the 11.4 copy outside any credential that production workloads hold.
- Backup encryption keys: 11.3 encryption only helps if the keys sit apart from the data. In the GoTo breach, an attacker stole encrypted backups from shared cloud storage and also exfiltrated an encryption key for part of them. Hold keys in a key management service or HSM with its own access policy, keep an out-of-band copy of the keys needed for recovery, and rotate after any nearby intrusion. See the Cryptographic Key Management Guide.
Backup console administrators should be separate accounts outside the production directory, with phishing-resistant MFA and, where the platform supports it, multi-person approval for deleting recovery points or changing retention. Recovery itself depends on identity: if Active Directory or Entra ID is down, plan how you will authenticate to the backup system. See the Identity Recovery Guide and the Break-Glass Account Guide.
Implementation steps
- Define scope and priorities (11.1): start from the Control 1 asset inventory and the Control 3 data inventory. Classify systems into recovery tiers with RPO and RTO for each, and include SaaS data, infrastructure-as-code templates and golden images, which CISA's #StopRansomware Guide recommends keeping as rebuild assets. See the CIS Control 3 Data Protection Guide.
- Write the process: cover scope, recovery order, roles, backup security (encryption, access, isolation), retention, how restores are requested and approved, and how backups are checked for malware before restore. Set an annual review date and a trigger for review on major change.
- Automate backups (11.2): configure jobs for every in-scope asset at weekly or better, with daily or continuous protection for high-tier data. Alert on failed and missed jobs; a backup that silently stopped is the most common audit finding.
- Protect recovery data (11.3): enable encryption at rest and in transit in the backup tool, store keys in a separate KMS or HSM, restrict console roles, and apply the same data access rules to restores as to the source.
- Isolate one copy (11.4): pilot immutability in a reversible mode (AWS governance mode, Azure immutable vault "Enabled"), confirm retention covers likely attacker dwell time, then lock it. Move backup administration into a separate identity domain.
- Test recovery (11.5): each quarter restore a sample across tiers to an isolated location, check application function and data integrity, and record time taken against RTO. At least annually, restore a tier 0 system such as a domain controller into an isolated network.
- Prove tamper resistance: as a planned test, attempt to delete a locked recovery point with an administrator account and confirm it fails and raises an alert.
- Rollback and change control: treat retention and lock changes as high-risk changes with approval. Before locking a vault, check that retention is right; AWS warns that recovery points with retention set to "Always" are kept forever once grace time expires.
Small versus large organisations
- Small (IG1): a SaaS backup service plus a locked cloud vault or rotated offline disk usually covers 11.2 to 11.4. Write the process in plain language and name who restores what. Make sure the backup service's admin account is not the same as your everyday Microsoft 365 or Google Workspace admin. See the SaaS Backup Buyer's Guide.
- Large (IG2/IG3): the hard problems are coverage drift, many backup products and proving recovery at scale. Reconcile backup job coverage against the inventory monthly, consolidate tools where possible, and run at least one full-service recovery exercise a year with the incident response team. See the Cloud Recovery and Clean Room Guide.
Evidence and assessment
The CIS Controls Assessment Specification measures Control 11 through the process document, counts of in-scope assets and their backup configuration, and dates.
| Safeguard | Evidence an assessor expects | Measure to report |
|---|---|---|
| 11.1 | Approved data recovery process with version history, covering scope, prioritisation and backup security | Elements present out of three; months since last update (fails above 12) |
| 11.2 | List of in-scope assets; backup job configuration; job history showing last successful backup per asset | In-scope assets with authorised, properly configured backup software; assets backed up within the last seven days |
| 11.3 | Backup encryption settings; key management configuration; console role assignments | In-scope assets whose backups are encrypted versus total in scope |
| 11.4 | Configuration showing the isolated destination (offline rotation log, separate account, locked vault settings such as AWS Locked and LockDate) | In-scope assets sending backups to an isolated instance versus total in scope |
| 11.5 | Quarterly test records: sample selected, restore location, result, time taken, issues and fixes | Backups tested, working and failing after restore; months between tests (fails above three) |
What to monitor
- Job health: failed, missed and partially completed jobs, and assets with no successful backup in their RPO window.
- Destructive actions: deletion of recovery points, retention or lifecycle changes, disabling of immutability, encryption setting changes and new console admins. In the cloud, alert on vault lock changes and bulk object overwrites.
- Host-level recovery inhibition: deletion of volume shadow copies and changes to boot recovery settings, both covered by T1490, are common ransomware precursors that endpoint tools under Control 10 should alert on.
- Backup identity use: backup service accounts logging on interactively or from unexpected hosts.
Metrics that show it is working
- Backup coverage: percentage of in-scope assets with a successful backup inside their RPO.
- Isolated coverage: percentage of in-scope assets with a copy in an immutable or offline destination.
- Restore success rate: percentage of sampled restores that worked first time, quarter by quarter.
- RTO achievement: measured restore time for each tier against its RTO, including one tier 0 system a year.
- Retention versus dwell time: oldest clean recovery point available compared with the dwell time you plan for.
- Standing backup admins: number of accounts able to delete recovery points or change retention.
Mapping to other frameworks
| Safeguards | NIST CSF 2.0 | NIST SP 800-53 Rev 5 | ISO/IEC 27001:2022 Annex A | ATT&CK techniques addressed |
|---|---|---|---|---|
| 11.1 | RC.RP-01, RC.RP-02 | CP-2, CP-10 | A.5.30 ICT readiness for business continuity | Recovery from T1486 Data Encrypted for Impact |
| 11.2 | PR.DS-11 | CP-9 | A.8.13 Information backup | T1485 Data Destruction; T1561 Disk Wipe |
| 11.3 | PR.DS-11, PR.DS-01 | CP-9(8), SC-28 | A.8.13; A.8.24 Use of cryptography | Backup theft and exposure |
| 11.4 | PR.DS-11, PR.IR-03 | CP-9(3), CP-6, CP-6(1) | A.8.13 | T1490 Inhibit System Recovery |
| 11.5 | PR.DS-11, RC.RP-03, ID.IM-02 | CP-9(1), CP-4 | A.8.13; A.5.30 | Confirms recovery from T1486 and T1485 |
NIST CSF 2.0 PR.DS-11 states that backups of data are created, protected, maintained and tested, which covers 11.2 to 11.5 in one outcome, and RC.RP-03 adds verifying the integrity of backups before using them for restoration. CIS publishes official v8.1 mappings; use those for formal crosswalks.
Common pitfalls
- Scope by habit: backups cover the servers that were set up years ago, while SaaS data, cloud databases and new file shares are missed. Derive scope from the inventory and reconcile monthly.
- Backup server in the production domain: one domain admin compromise gives the attacker the backup console and the means to delete every recovery point. Move backup administration into its own identity domain.
- Governance mode treated as immutable: a reversible lock stops accidents, not an attacker with the right role. Lock at least one copy in compliance or locked mode once retention is agreed.
- Keys stored with the data: encryption meets the 11.3 checkbox but not its intent when the key travels with the backup, as the GoTo breach showed. Separate them.
- Long-lived cloud keys with write and lifecycle rights: the Codefinger campaign needed nothing more. Use short-lived roles and keep the isolated copy outside their reach.
- Testing that never restores: checking that jobs completed is not 11.5. Restore real samples, check they work, and record the time against RTO.
Practitioner checklist
- Confirm the target IG: IG1 needs 11.1 to 11.4; IG2 and IG3 add quarterly recovery testing (11.5).
- Approve a data recovery process covering scope, recovery prioritisation with RPO and RTO, and backup security; review it at least annually.
- Back up every in-scope asset automatically at least weekly, more often for sensitive and fast-changing data, and alert on failed jobs.
- Encrypt backups, hold keys in a separate KMS or HSM with an out-of-band recovery copy, and restrict restore access to match the source data.
- Keep at least one isolated copy in offline media or a locked, immutable vault in a separate account or identity domain.
- Scope backup service accounts and cloud keys to append-only, deny deletion and retention changes, and vault their credentials.
- Alert on recovery point deletion, retention and lock changes, shadow copy deletion and new backup admins.
- Restore a sample each quarter and a tier 0 system each year into an isolated network, scanning for malware before restore.
- Report backup coverage, isolated coverage, restore success and RTO achievement to the risk owner.
Standards and references
- CIS: Control 11 Data Recovery
- CIS: Controls Assessment Specification, Control 11
- CIS: CIS Controls v8.1
- NIST: SP 800-34 Rev. 1 Contingency Planning Guide for Federal Information Systems
- NCSC: Principles for ransomware-resistant cloud backups
- CISA: #StopRansomware Guide
- AWS: AWS Backup Vault Lock
- Microsoft Learn: Immutable vault for Azure Backup
Related NHI Mgmt Group resources: CIS Controls Guide · Backup Security and Immutable Backup Guide · CIS Control 10 Malware Defenses Guide · Ransomware Response and Recovery Guide · CIS Control 17 Incident Response Management Guide