Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Recovery is the phase that returns systems to service after an incident, under heightened monitoring and only once the attacker’s access has genuinely been removed.

What Is Recovery in Incident Response?
Recovery is the phase of incident response in which systems return to normal operation. It sounds like the easy part and is usually the longest, most expensive, and most politically difficult. Containment and eradication are measured in hours and days; recovery from a destructive attack is measured in weeks and months.
The defining constraint is that you cannot start it honestly until you can answer one question: is the attacker out? Bringing services back while an implant remains gives the attacker a restored, patched, repopulated environment to work in, and the second incident is invariably worse than the first.
Restoring from backup is faster. Rebuilding from known-good images is safer. The choice turns on confidence: if you cannot establish when the attacker first gained access, you cannot know whether a backup predates the compromise, and restoring a backdoored system is a real outcome rather than a theoretical one.
In practice most organisations split the decision. Domain controllers, identity infrastructure, and anything the attacker demonstrably touched get rebuilt. Bulk file data and databases are restored, after scanning, into a clean platform. Where backups are the only option and their integrity is uncertain, restore into an isolated environment and examine before connecting anything.
Rebuilding servers achieves little if the attacker still holds valid credentials. A thorough rotation covers more than user passwords:
krbtgt account, reset twice with sufficient interval between resets for replication to complete, to invalidate forged Kerberos tickets.This is also the moment to remove the access paths that were abused: the flat network, the SMB exposure between workstations, the unmonitored remote access server.
Bring services back in a planned order driven by business criticality and risk, not by whoever is shouting loudest. Each wave should be patched, hardened, and freshly credentialed before it reconnects, and the environment should run under enhanced logging and alerting for a defined period afterwards, commonly thirty to ninety days, with the incident response provider retained throughout.
Set re-entry detection deliberately. If you know which tooling and infrastructure the attacker used, alert specifically on it. A quiet fortnight is not evidence of success; it is the normal dwell pattern of an actor waiting for the response team to stand down.
The British Library published an unusually candid account of its October 2023 Rhysida ransomware attack. Its recovery was shaped less by the ransomware than by what the ransomware exposed: legacy systems that could not simply be restored because the underlying infrastructure no longer existed in a supportable form, and a complex estate without a complete picture of its own dependencies. The main catalogue returned in read-only form in January 2024, and the institution estimated that recovery would consume a substantial share of its reserves.
The report’s central lesson is that recoverability is an architecture problem decided years earlier, not a decision made during the incident. Legacy technology, incomplete asset knowledge, and backups that have never been restored at scale all convert a bad week into a bad year.
Recovery ends on evidence, not on fatigue. Reasonable exit criteria include: all identified persistence mechanisms removed and verified; credentials rotated and the rotation confirmed; patching of the exploited vulnerability complete across the estate; monitoring restored and tuned; the root cause established and the corresponding control implemented; and documented business sign-off that services meet agreed performance.
Test the recovery capability before you need it. Restoring a single file proves nothing about restoring a domain. Our Cyber Exercise Services put that assumption under pressure in a controlled setting, which is considerably cheaper than discovering the answer during a real incident.
Eradication removes the attacker’s access and tooling: deleting implants, closing the entry point, disabling compromised accounts. Recovery returns systems to productive service afterwards, through rebuilding or restoring, revalidating data, and reconnecting under heightened monitoring. The two overlap in practice, and recovery should not begin for any system until eradication has been verified for the environment it will rejoin.
By establishing the earliest confirmed attacker activity and selecting a restore point before it, then verifying rather than assuming. Restore into an isolated environment first, scan for known indicators, check for unexpected scheduled tasks, services, and accounts, and compare system binaries against known-good versions. Immutable or offline backups are far easier to trust because the attacker could not have altered them.
Active Directory keeps the current and previous password for the krbtgt account, so a single reset leaves forged Kerberos tickets signed with the old key still valid. Resetting twice invalidates both. The two resets must be separated by enough time for replication to complete across all domain controllers, otherwise authentication breaks for users served by a lagging controller.
Thirty to ninety days is the common range, extended where the actor is sophisticated or the entry point was never conclusively identified. Keep enhanced logging, retain the incident response retainer, and alert specifically on the tooling and infrastructure observed during the incident. Attackers frequently wait for the response team to disengage before returning through a foothold that was missed.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.