Skip to content
Knowledge base

Clean restore point after ransomware: how to check a backup before you restore it

Servers are encrypted, last night's backup looks untouched and the board is asking when the company will be back at work. Before anyone clicks "restore", you need to establish whether that backup contains the same attacker. A clean restore point is set by the timeline of the intrusion, not by the date of the last successful backup.

Monday morning. The file servers and the ERP system are encrypted, the backup console shows a green job from last night, and the board wants to know when the company will be back at work. The administrator has the cursor over the “restore” button. The domain controller logs, however, show a service account that nobody created, set up long before the encryption.

This is the moment when ransomware recovery is easiest to get wrong. A backup that the ransomware did not encrypt is not yet a clean backup: it may have been taken while the attacker already had accounts, tools and scheduled tasks in the network. Recovery testing, the first hours of an incident and reporting to CERT Polska are covered in the guide to ransomware protection.

What is a clean restore point after a ransomware attack?

A clean restore point after a ransomware attack is a backup of a system or data taken before the attacker entered the network, free of their tools and persistence mechanisms, with data whose integrity has been checked. All three conditions have to be met at once. A backup taken before the encryption, but during the period when the intruder was already active, meets only some of them.

NIST describes this goal in SP 1800-11 “Data Integrity: Recovering from Ransomware and Other Destructive Events”, published in September 2020. Among the guide’s goals it lists restoring data to its last known good configuration and identifying the correct backup version, free of malicious code and data.

Why might the latest backup not be safe?

The latest backup might not be safe because encryption can be the last stage of an attack rather than the first. The #StopRansomware Guide, developed by the US organizations CISA, MS-ISAC, NSA and FBI (September 2023 version), notes that a ransomware event may be evidence of a previous, unresolved network compromise. Operators of malware that prepares the ground often sell access to the network to other groups.

This leads to a specific CISA warning: dropper malware has to be identified before rebuilding from backups, to prevent continuing compromise. A backup taken while the dropper was already running on the server will carry it into the restored environment together with the data. A green backup job status says nothing about it.

The second problem concerns the backups themselves. Among the signs to check, CISA lists anomalous use of built-in Windows tools such as vssadmin.exe, wbadmin.exe or bcdedit.exe, describing their misuse as a common technique to inhibit system recovery. An attacker who deleted shadow copies may also have changed the schedule or retention of backup jobs. That is why the list of available restore points has to be compared with what should be on it.

Key takeaways:

  • A backup from before the encryption is not automatically a backup from before the intrusion.
  • CISA warns that dropper malware must be found before rebuilding from backups, otherwise the compromise continues.
  • The list of available backups may have been changed by the attacker — compare it with the schedule and retention.

How do you establish when the attacker entered the network?

The attacker’s point of entry is established from the incident timeline, built from logs, EDR data and traces of changes to accounts and configuration. The earliest credible trace of intruder activity sets the boundary: backups newer than it are suspect, older ones are candidates for restoration. In SP 1800-11, NIST points to logs as exactly the tool that helps identify the time of the last known good state and inform the restoration decision.

In the detection and analysis part of its guide, CISA provides a list of things to check during threat hunting. It includes newly created Active Directory accounts and accounts with escalated privileges, anomalous VPN logins, unexpected use of remote monitoring and management (RMM) software, and new services and scheduled tasks. Each of these findings has a date, and the earliest of them moves the clean point boundary back.

The timeline may be incomplete, because logs have limited retention and some systems were not monitored. In that case the boundary is set from the earliest confirmed trace, with a margin for gaps in the logs, and recorded together with the rationale.

Before anyone starts restoring, it is worth preserving the material for this analysis. CISA recommends taking a system image and memory capture from a sample of affected devices and, for cloud resources, snapshots of volumes for later forensic review. Restoring onto the same hardware without such preservation erases the traces that define the clean point.

How do you choose a restore point when systems and data have different dates?

Systems and data need separate decisions: operating systems and applications are best rebuilt from clean images, while data is restored from a backup selected on the basis of the timeline. CISA recommends rebuilding systems based on the prioritisation of critical services, using preconfigured standard images, and rebuilding cloud resources from infrastructure as code templates. CERT Polska’s “Ransomware” guide describes the same step: restoring systems, starting with those critical to the business, using clean, preconfigured images.

Separating the two paths has a simple reason. Malicious code and persistence mechanisms live mainly in the system: in services, tasks, the registry and local accounts. Business data, such as documents or database records, may also have been altered, but it is checked with different methods. Restoring an entire virtual machine from a single backup carries both types of risk at once.

Choosing the data backup is a trade-off between data loss and risk. An older backup is safer, but it means losing transactions, documents and changes from a longer period. How much data the company can afford to lose is decided by the business process owner, not the administrator — the technical team gives them the boundary from the timeline and the result of the backup verification. Encrypted files should not be deleted in the meantime: CERT Polska recommends keeping a copy of them in case a decryptor appears for the ransomware family involved.

Where should a backup be checked before it is restored?

A backup is checked in an environment isolated from production, with a separate network and separate credentials, before anything returns to the corporate network. In SP 800-184 “Guide for Cybersecurity Event Recovery”, published in December 2016, NIST lists among the resources needed for recovery a staging system to validate the integrity of data recovered from backups, in order to re-establish trust in the infrastructure.

CISA adds a rule for the recovery network itself: if a new VLAN has been created for recovery purposes, only clean systems may be added to it.

Isolation also applies to identity. If the recovery environment authenticates administrators against the directory the attacker has taken over, the intruder has access to it from the start. That is why accounts created specifically for this purpose are used there, and administrators’ workstations from production are not connected to it.

What should you check in a restored backup before it goes into production?

Two things are checked in a restored backup: whether it contains traces of the attacker and whether the data is complete and consistent. NIST describes the second of them, integrity, in the profile IR 8374r1 “Ransomware Risk Management: A Cybersecurity Framework 2.0 Community Profile”, published in June 2026 (it superseded the 2022 version). Outcome RC.RP-03 reads: the integrity of backups and other restoration assets is verified before using them for restoration.

Traces of the attacker are searched for using the indicators from the investigation — file hashes, service names, scheduled tasks, accounts, registry keys — and the CISA threat hunting list. An antivirus scan alone is not enough, because CISA describes attackers maintaining access with legitimate tools: RMM software, the PsTools suite or PowerShell scripts.

Data consistency is checked by the application, not only by the file system. A database should start and pass its own integrity checks, the ERP system should open the last closed periods, and files should open in the programs that use them. In SP 1800-11, NIST describes a separate corruption testing component that makes it possible to determine when files were modified and thus identify the last known good state.

The verification result is recorded separately for each backup: what was checked, with what, by whom and with what result. This is the basis for the decision to return to production and, where personal data is involved, also evidence of the ability to restore the availability of data in a timely manner and of regular testing of measures, which Article 32(1)(c) and (d) of the GDPR requires (as of 16 November 2026).

How do you restore identity and access without letting the attacker back in?

Identity is restored after the attacker has been cut off from access channels and the environment has been cleaned, not before. Accounts and systems used by the attacker are identified and cut off already during containment: CISA recommends identifying the systems and accounts involved in the initial breach, including email accounts, and disabling VPNs, remote access servers, single sign-on (SSO) resources, and cloud-based or other public-facing assets if they could be used for continued unauthorised access. It also notes that breaches often involve mass credential exfiltration.

The CISA guide gives mass password resets a clear place in the sequence: after the environment has been fully cleaned and rebuilt, together with removing persistence mechanisms, patching vulnerabilities and updating customer-managed encryption keys. A reset carried out earlier, while the intruder still has access to the domain controller, may simply give them the new passwords.

The user directory needs the most attention. An Active Directory backup from the period after the attacker’s entry contains their accounts and permissions, so restoring it without review gives them their access back. The list of privileged accounts, administrative groups and service accounts in the restored backup should be compared with the expected list before the directory starts serving production.

In what order should systems be returned to production?

Systems are restored according to a predefined list of critical services, only after the incident has been formally declared over and only from verified backups. CISA recommends triaging systems for restoration on a clean network according to a critical asset list covering systems critical for health and safety, revenue generation, and the systems they depend on.

Before systems return to the network, the designated IT or IT security authority declares the incident over based on established criteria. Only in the recovery phase does CISA recommend reconnecting systems and restoring data from offline, encrypted backups, again according to service priority, taking care not to reinfect clean systems.

The order is also set by dependencies. A business application needs identity, DNS, a database and the network, so restoring it before them ends in a series of failed start-ups. A dependency map should be part of the plan prepared before the incident.

Key takeaways:

  • Rebuild systems from clean images and restore data from the backup selected after timeline analysis.
  • Backups are verified in an isolated environment, with separate administrative accounts.
  • Cut off accounts used in the intrusion already during containment, and carry out the mass password reset after the environment has been cleaned.

What should you prepare before an attack so that a clean point exists at all?

A clean restore point exists only if the backups survived the attack, reach far enough back and are stored out of the attacker’s reach. CISA recommends offline, encrypted backups of critical data, with regular testing of their availability and integrity, because many ransomware variants attempt to find and delete or encrypt accessible backups. Outcome PR.DS-11 requires backups to be created, protected, maintained and tested, and the NIST IR 8374r1 profile adds in its commentary that one copy should be stored offline or otherwise out of the attacker’s reach.

The second part of the preparation is system images. CISA recommends maintaining and regularly updating “golden images” of critical systems and keeping infrastructure as code templates offline — without them, rebuilding from a clean image turns into manual installation.

Immutable backups help, but they do not solve the cleanliness problem. They protect a backup from deletion, but they do not tell you whether it was free of malicious code when it was taken. CISA also urges caution: immutable storage does not meet compliance criteria for certain regulations, and misconfiguring it can impose significant cost. We break down the mechanism and its limits in the article on immutable backup.

Retention should follow from how long an attack can stay unnoticed in a given company, not from the default settings of the backup system. Backup types, the 3-2-1-1-0 rule and storage locations are covered in the guide to backup and disaster recovery; what matters here is one point: CERT Polska’s “Ransomware” guide stresses that at least one copy should be isolated and that backups must be regularly tested for restoration, so that an error in a backup does not surface only after an infection.

How do you assess a company’s readiness for a clean recovery?

Readiness is assessed by evidence that can be shown before an incident, not by declarations. The table below collects control questions that follow from the previous sections; each has the evidence that settles it and a sign that the answer is worse than it seems.

Control questionEvidence that settles itWarning signWho is responsible
Can we establish when the attacker entered?retention of AD, VPN, EDR and backup logs and where they are storedlogs kept on servers that can be encryptedIT security
Do we have backups from before that boundary?backup retention compared with log retentionretention shorter than the period covered by logsbackup administrator
Are backups out of the attacker’s reach?offline or isolated copy with separate authenticationbackup console accessed with a domain accountbackup administrator
Do we have clean images of critical systems?current golden images and IaC templates stored offlineimages older than the current system versionssystem administrators
Do we have somewhere to verify backups?a documented and tested isolated environmentthe plan assumes restoring straight into productioninfrastructure architect
Do we know the order of return?list of critical services with a dependency mappriorities set only during the incidentprocess owners and IT
Does the data loss decision have an owner?a named person in the response planthe administrator picks the backup alonemanagement board
Has the procedure been exercised?record of the last restore with isolation and verificationthe test covered only file restorationIT manager

The table shows where an ordinary restore test gives a falsely good result: restoring yesterday’s backup in a clean lab does not check whether the company could pick a backup from before the intrusion. More material is collected in our ransomware hub.

How can nFlo help you prepare for a clean recovery after ransomware?

nFlo helps prepare recovery before it is needed: as part of the backup and disaster recovery service, we design backup policies, retention and recovery tests, and we agree RTO and RPO for each system in the contract. We implement Veeam, Commvault and IBM Storage Protect.

Where a copy cut off from production is needed, we implement Dell PowerProtect with Cyber Recovery Vault, an isolated environment for offline copies. For SAP HANA, Oracle and Epic systems, we implement IBM Storage Defender Sentinel (formerly IBM Storage Sentinel), which analyses the content of snapshots and helps determine which copy is the last uncorrupted one. None of these tools replaces timeline analysis or the verification described above. nFlo carries out that analysis, examining the traces of the attack (digital forensics) and reconstructing its course, as part of its post-incident management service.

If you want to check whether your company could identify a clean restore point, let’s talk.

What else is worth knowing about ransomware recovery?

Can you restore the latest backup after a ransomware attack?

Only if the incident analysis shows that it was created before the attacker entered the network. CISA warns that dropper malware has to be identified before rebuilding from backups, otherwise the compromise continues.

What is the difference between rebuilding a system and restoring data from a backup?

Rebuilding means setting the system up from scratch using a clean, preconfigured image, as both CISA and CERT Polska recommend. Restoring data means moving files or databases from the backup selected after analysis into the rebuilt system.

Where should a backup be checked before it is restored?

In an environment isolated from production, with separate credentials. NIST SP 800-184 lists a staging system for validating the integrity of data recovered from backups, and CISA recommends that only clean systems are added to the recovery network.

Is an immutable backup enough to make recovery safe?

No. Immutability protects a backup from deletion and encryption, but it does not tell you whether the backup was clean when it was taken.

When should passwords be changed after a ransomware attack?

The CISA guide places mass password resets after the environment has been fully cleaned and rebuilt. Earlier, already during containment, the accounts and systems used in the intrusion are identified and cut off, and VPNs, remote access servers and single sign-on resources are disabled.

Share:

Talk to an expert

Have questions about this topic? Get in touch with our specialist.

Sales Representative
Grzegorz Gnych

Grzegorz Gnych

Sales Representative

Response within 24 hours
Expert consultation
Individual approach

Providing your phone number will speed up contact.

Want to Reduce IT Risk and Costs?

Book a consultation - we respond within 24h

Response in 24h Custom quote No obligations