Research · Explainer

Backups That Survive Ransomware: Offline Copies, Golden Images and Tested Restores

Attackers hunt reachable backups first. The federal ransomware guide’s preparation section is built around making sure the copies that matter are the ones they cannot touch, and that restoring them is a rehearsed procedure, not a hope.

A 3.5 inch desktop hard disk drive with PATA connections.
An external hard drive cabled to a laptop. The guide’s first backup rule is physical as much as digital: at least one copy of critical data has to sit where an infected network cannot reach it.

Photo: Arina Krasnikova, via Pexels

Every ransomware negotiation starts from the same question: can the victim restore without paying? The joint #StopRansomware Guide, written by CISA with the Multi-State Information Sharing and Analysis Center, the National Security Agency and the FBI, answers it in its preparation section, before any discussion of the incident itself. Its core instruction is to maintain offline, encrypted backups of critical data, and to test them. Everything else in the guide’s backup material exists to defend that sentence against the ways backups fail in practice.

Offline is the point, not a preference

The guide gives the reason plainly: many ransomware variants actively hunt for backups that the infected systems can reach, and delete or encrypt them, so that restoration without payment becomes impossible. A backup that a compromised server can write to is a backup the ransomware can write to. Offline copies, whether physically disconnected media, a separate environment, or immutable cloud storage that cannot be altered inside its retention window, break that reach. The guide adds a caution on immutable cloud storage: use it carefully, because it does not meet the compliance criteria for certain regulations and a misconfiguration can impose significant cost. It also suggests considering a multi-cloud arrangement for cloud to cloud backups, so that one vendor’s compromised accounts do not take every copy with them.

A Hitachi 2.5 inch laptop hard disk drive.
Connecting an external drive to a laptop. The guide’s harder requirements sit behind the simple copy: separation from the network, retained backup hardware for rebuilds, and restores that are actually rehearsed.

Photo: Andrea Piacquadio, via Pexels

Back up the rebuild, not just the data

Restoring files onto broken systems is the slow path. The guide recommends maintaining and regularly updating golden images of critical systems: templates with a preconfigured operating system and the associated software applications, ready to deploy to rebuild a machine, a virtual machine or a server. Where infrastructure is deployed as code, the templates themselves belong in the backup set, version controlled, audited when they change, and copied offline so the environment can be redeployed quickly. Source code and executables should be stored with the offline backups, along with escrowed materials and license agreements, because an image that will not install on replacement hardware is a delay, not a recovery. The guide even covers the hardware: retain backup hardware for rebuilds where rebuilding the primary system is not preferred, and replace aging hardware that would obstruct restoration before an incident forces the question.

A backup you have not restored is a hypothesis

The guide’s testing requirement is specific: put the copies through complete restore exercises on a recurring schedule, run under disaster recovery conditions, so that gaps surface while they can still be fixed. That means timed, full restores of real systems from the offline copies, including the golden images, run when nothing is on fire, with the failures fixed and the run repeated. It also means knowing, in numbers, how long a restore takes, because that number is the organization’s real downtime floor in a ransomware event and the figure that belongs in front of leadership before an incident, not during one. Pair the restore tests with the guide’s other preparation step, a maintained and exercised incident response plan with an offline copy available, and the backup program stops being a storage purchase and becomes what it is supposed to be: the reason the ransom demand is a problem to manage rather than a bill to pay.

No corrections have been published for this article.

About the author

Graham Ellis, Staff Writer, beat: consumer security & privacy.