Cybersecurity

Ransomware-Proof Backups: How to Keep One Copy Out of Reach

A backup only helps if ransomware cannot reach it. Using guidance from the UK's NCSC and the US CISA, here is how the 3-2-1 rule, offline copies and restore testing fit together.

Silver and black hard disk drive

A backup protects you from ransomware only if the attackers cannot reach it. Guidance from the UK National Cyber Security Centre (NCSC) and the US Cybersecurity and Infrastructure Security Agency (CISA) comes down to three habits: keep several copies in separate places, keep at least one disconnected from everything else, and test that you can restore from it.

Why ordinary backups often fail against ransomware

Ransomware is malicious software that locks or destroys data and demands payment. Our explainer on how ransomware attacks unfold covers the broader picture. The backup problem is narrower and more practical.

According to CISA’s #StopRansomware Guide, many ransomware variants try to find and then delete or encrypt backups they can reach, which can make restoration impossible unless the ransom is paid. The NCSC makes a similar point: early in a destructive attack, criminals often go after backups and the infrastructure behind them, because destroying them makes recovery harder and payment more likely.

The NCSC has also said it has seen ransomware encrypt not only live data but also connected USB, network and cloud storage holding backups. A backup drive that stays plugged in, or a shared network folder that every computer can write to, is part of the same environment the attacker is already inside. The NCSC adds that backups are not resistant to ransomware by default.

What is the 3-2-1 backup rule?

The 3-2-1 rule is the most common starting point. As the NCSC describes it, keep at least three copies of your data, on two devices, with one offsite. The offsite copy can be in the cloud.

Part of the rule What it means in practice
3 copies The live data plus two backups, so one failure does not end recovery
2 devices Different kinds of storage, so a single hardware fault or a single compromised system cannot take everything
1 offsite A copy away from the main location, which can be a cloud service

The NCSC’s reasoning is that multiple backups in logically separated locations mean that if one is compromised, at least one other remains. The rule is useful for fires, theft and hardware failure. Against ransomware it has a gap, which the NCSC states directly: 3-2-1 does not require any copy to be offline.

Why does one copy need to be offline?

This is the “offline rule.” The NCSC’s aim is that at least one backup stays unaffected by an incident in the live environment, so it advises never having all backups connected, or “hot,” at the same time. CISA’s guide likewise says to maintain offline, encrypted backups of critical data.

For a home user or small business with a physical drive, that can mean a drive that is plugged in only while the backup runs, then unplugged and stored. Disconnecting the drive after each run is the simplest form of the offline rule. Rotating between two drives, one connected and one stored, is a common extension.

What “offline” means for cloud storage

Cloud storage cannot be unplugged, so the NCSC describes a digital version of the same idea, built from three measures:

  • Identity management: protect the accounts that control the backups with strong passwords and multi-factor authentication, so ransomware cannot simply request access. Our comparison of multi-factor authentication methods explains the options.
  • Client management: backup software on a device should not hold valid credentials while the storage is not in use. Keep the number of backup clients small and stop ordinary user devices from modifying backups directly.
  • Access control: where the service allows it, let only authorized clients create or add backups, refuse connections while the storage is idle, and use permissions that can add data but not overwrite existing backups. The NCSC compares this to write-once storage.

CISA’s guide recommends enabling delete protection or object lock, meaning settings that stop stored data from being removed or changed for a period, on storage that ransomware often targets. It also advises caution with immutable storage, because it may not meet some regulatory compliance criteria and a misconfiguration can be costly.

How often should backups run? A worked example

The NCSC’s fourth principle, the “regular rule,” says to back up critical data on a schedule: more frequent backups mean less data lost during recovery. The tradeoff can be shown with simple, illustrative arithmetic.

Suppose a small office runs one backup every night at 2:00 a.m. Ransomware locks the files at 5:00 p.m. the next day. The most recent clean backup is 15 hours old (2:00 a.m. to 5:00 p.m.), so up to 15 hours of work is gone. In the worst case, an attack just before the next run, the gap is the full 24 hours. If the same office backed up every 4 hours instead, the worst-case loss would fall to 4 hours.

The numbers are invented for illustration. The principle is the point: the interval between backups sets the most work you can lose. A faster schedule also creates more copies to protect, so it works best when the offline and access-control measures above still apply.

Is a backup you have never restored really a backup?

CISA says to regularly test the availability and integrity of backups in a disaster recovery scenario, meaning a realistic rehearsal of rebuilding after a serious failure. The NCSC says the same about cloud data: test whether it can actually be restored.

Testing finds problems that are invisible until the worst moment:

  • backups that run but capture the wrong folders;
  • files that are corrupted or incomplete;
  • restores that take far longer than anyone planned;
  • recovery steps that depend on a password stored on the very system that has been encrypted.

For individuals, a basic test is to restore a handful of files, open them, and confirm they are intact. Organizations need a fuller rehearsal.

What else makes recovery possible?

Data is only part of recovery. CISA’s guide also advises:

  • Golden images: keep and regularly update preconfigured templates of critical systems so they can be rebuilt quickly.
  • Infrastructure as code: keep the template files that describe cloud resources offline so those resources can be redeployed.
  • Source code and executables: store them with the offline backups.
  • Spare hardware: keep hardware for rebuilding systems, and replace outdated equipment that could complicate restoration.
  • Cloud redundancy: consider a multi-cloud approach for cloud-to-cloud backups to avoid being locked into one vendor, and back up often.
  • Third parties: if a provider maintains your backups, make sure it follows these practices and write the requirements into the contract.

The NCSC also lets buyers ask whether a backup service is resistant to destruction. It has published two sets of principles for this, one for on-premises backups and one for cloud-based services. They are meant for vendors to show how their service meets them and for system owners to review a solution or prepare questions for a supplier.

Common mistakes and misconceptions

“The cloud is automatically a backup.” A synced folder can mirror a deletion or an encryption event. A real backup keeps earlier versions. The NCSC advises choosing providers that can restore modified data to an earlier version and recover deleted data for a limited time, so you can return to the last known-good state.

“Three copies on the same network is 3-2-1.” The count matters less than the separation. If one set of credentials reaches every copy, one intruder can reach them all.

“Backups solve ransomware.” They address destruction, not theft. The NCSC’s principles state that they address destructive attacks, not data theft for extortion. For that risk, the backup system should be protected from unauthorized access like any other system holding sensitive data. Prevention still matters, including patching, a topic we cover in our piece on risk-based patching and the CISA KEV catalog.

“Restoring is the last step.” CISA’s guide warns against reinfecting clean systems during recovery. It advises identifying dropper malware, the program that installs further malicious code, before rebuilding from backups, and restoring in order of the most critical services.

What varies by situation

A household, a clinic and a software company need different setups. A household may rely on a rotated external drive and a cloud service protected by multi-factor authentication. An organization with regulatory duties has to weigh immutable storage against compliance rules, as CISA cautions. The common thread is the same: at least one recoverable copy that an intruder in the main network cannot touch, and proof that it works.

A short checklist to start from

A reasonable first pass, drawn from the NCSC and CISA guidance, is to list the data you cannot lose, make at least three copies on two kinds of storage with one offsite, keep one copy disconnected or digitally locked down, protect backup accounts with multi-factor authentication, run backups on a schedule that matches how much loss you can tolerate, and test a restore. Repeat the test periodically, because systems change.

This article is general information, not a substitute for professional security advice.

Sources: National Cyber Security Centre; Cybersecurity and Infrastructure Security Agency