Skip to main content

Hit by ransomware? Isolate affected systems now. Do not reboot or reformat.

SheMo Noransom舍末无勒

Scenario solution

Backups Deleted or Destroyed

  • Veeam Backup & Replication
  • Windows Server Backup
  • Commvault / NetBackup
  • NAS 备份目标
  • 磁带库
  • 卷影副本 VSS
  • 云对象存储

Modern ransomware follows a fixed sequence: destroy the backups, then encrypt the data — deleting shadow copies, encrypting repositories, disabling jobs, and exploiting backup software flaws to steal credentials. This page covers what can still be inventoried once backups fail, why replication propagates encrypted files off-site, and what offline and immutable copies are really worth.

Typical symptoms

  • Every job in the backup console fails, and the repository shows as inaccessible or reports abnormal capacity
  • Backup files on platforms such as Veeam (.vbk / .vib / .vbm) carry appended extensions or have been deleted, leaving an empty restore point list
  • Previous Versions is empty, shadow copies have been cleared, and vssadmin delete shadows activity appears in the logs
  • The backup directory on a NAS now contains encrypted files, and the off-site replication job has dutifully synchronised them
  • The backup server itself has been encrypted or logged into, with backup administrator accounts disabled, password-changed or newly created
  • Tape jobs have been cancelled, retention shortened, or versioning and lock policies changed on an object storage bucket

Business risk & common mistakes

"We have backups" and "we can recover" are two different statements. In real incident response, the reasons backups fail cluster tightly into a handful of categories.

First: the backup target sat inside the attacker's permission scope. Backups written to an SMB share, a NAS share, or another partition on the same host that the production server could write to. As the encryptor enumerates network shares it follows write permissions outward, and the backup files fall squarely inside. This is far and away the most common cause.

Second: backup credentials were stolen. The backup server is domain-joined and runs jobs under a domain administrator account, so a compromised domain controller takes the backup estate with it. Further, public reporting documents ransomware operators exploiting flaws in backup software itself — for example the credential-disclosure vulnerability in Veeam Backup & Replication (CVE-2023-27532), which allowed unauthenticated retrieval of encrypted credentials from the configuration database and their decryption for lateral movement. Backup software is part of the attack surface.

Third: replication carried the encrypted files away. A recurring pain point: the NAS or cloud drive is configured for synchronisation rather than versioned backup, so after encryption the sync job faithfully pushes the encrypted files off-site and overwrites the only good copy. Sync is not backup — without version retention, sync is close to worthless in a ransomware scenario.

Fourth: shadow copies were cleared. Virtually every major family deletes shadow copies, usually before encryption begins, so VSS cannot be relied upon.

Fifth: the backups were never verified. Jobs reported success and the files exist, but no real restore was ever performed — and only when needed does the broken chain, incomplete version set or unreadable data come to light.

A caution: do not delete the encrypted backup files to free space once you discover the failure; they may be header-only encrypted and still repairable. Nor should you reinitialise the repository. We do not recommend paying, and we do not negotiate on your behalf.

Our response plan

  1. Stop all sync and overwrite activity first

    The first priority is not recovery but stopping the situation from getting worse. Immediately pause every backup and synchronisation job — NAS sync, cloud drive sync, replication tasks, tape rotation — and disconnect the backup server from the production network, so encrypted files stop propagating to off-site copies and retention policies do not expire the last good version. Preserve the backup server as-is for forensics: no reinstall, no repository reinitialisation.

  2. Inventory every copy that still exists

    Inventory by one criterion — whether the attacker could reach it. Cover offline media (tape, periodically detached external drives, archived optical media); immutable storage (object lock / WORM, repositories with immutability enabled); storage and hypervisor snapshots (array snapshots, read-only NAS snapshots, VM snapshots); off-site copies (secondary data centre, cloud, third-party custody); logical copies (database standbys, downstream warehouses, interface landing files); and distributed copies on endpoints and sync services. Many organisations discover at this step that usable resources do exist.

  3. Assess whether damaged backups can be repaired

    An encrypted backup file is not automatically worthless. Backup files are typically enormous — tens of gigabytes to several terabytes — and many families encrypt only the header of large files or use intermittent encryption, so the untouched portions may still yield complete backup data blocks. The method is segmented entropy analysis of .vbk, .bak, .tib and similar files; once coverage is established, rebuild the header or extract content directly from the unencrypted segments. Deleted backup files can also be carved while the storage has not been heavily overwritten.

  4. Build a recovery plan from what is usable

    Combine the inventory with the recoverability of production data and build a composite plan rather than a single path: restore from offline or immutable copies to an earlier point, close the gap using database logs, downstream systems and interface data, and fill in what the backup did not cover by repairing encrypted files on the production side. Agree priorities and an acceptable data gap with the business, and state explicitly which data will not come back so re-entry can be planned in advance.

  5. Rebuild a ransomware-resistant backup architecture

    After recovery, the backup architecture must be rebuilt or the next incident ends the same way. The core principles: the backup target must not be writable from production, the backup estate must not share credentials with production, and at least one copy must be offline or immutable. In practice: take the backup server out of the production domain with its own accounts and MFA; enable immutable backups or object lock; maintain an offline media rotation; patch the backup software itself and restrict access to its management ports; and validate RTO/RPO with real restore drills rather than job success rates.

Recovery paths

Once backups have failed, the work becomes finding copies the attacker could not reach, and extracting as much as possible from the wreckage.

Path 1: genuinely offline copies. Tape, periodically detached external drives, archived optical media, media held in a safe. Their value in ransomware scenarios is badly underrated — they were simply not online when the attack happened, so the attacker could not touch them. Even if the data only runs to last month, it is a certain, usable starting point.

Path 2: immutable backups and object lock. Repositories with immutability flags, cloud buckets with object lock, WORM-retained snapshots — none deletable within the retention period even with compromised administrator credentials. This is the central direction of recent backup architecture work.

Path 3: storage and hypervisor snapshots. Array snapshots, read-only NAS snapshots and VM snapshots generally sit outside the permission scope of the encrypted filesystem, and survive far more often than shadow copies.

Path 4: repairing encrypted backup files. Backup files are large and are often only header- or segment-encrypted. Once entropy analysis establishes coverage, headers can be rebuilt and data blocks extracted from unencrypted regions. The result may be some restore points usable or some VMs or databases extractable rather than an intact chain — but where other paths have failed, that is worth a great deal.

Path 5: carving deleted backup files. Deletion only releases metadata references, so files can be recovered while the storage has not been heavily overwritten. Again, this depends on writes stopping immediately.

Path 6: logical and distributed copies. Physical database standbys, downstream data warehouses, reporting systems, interface reconciliation files, third-party platform data, and version history on endpoints and sync services. None of these is a backup in the traditional sense, yet in practice they frequently do the heavy lifting in restoring critical business data.

Stated plainly: where every copy was online and writable, sync overwrote the off-site version, and the storage continued to be written after the incident, very little room remains. We avoid "100%" and "guaranteed recovery"; the assessment states what can and cannot be retrieved. No ransom payment, no negotiation.

Common ransomware families

Hardening recommendations

  • 3-2-1-1 is the floor, not the aspiration. Three copies, two media types, one off-site, and at least one offline or immutable. That final copy is the one that actually helps in a ransomware scenario.
  • The backup target must never be writable from production. Do not write backups to an SMB or NAS share the production server can reach; use a dedicated backup host, separate storage, an object-locked cloud bucket, or tape.
  • Decouple the backup estate from production. Keep the backup server out of the production domain, with its own local accounts, strong passwords and MFA, and separate backup and production networks. A compromised domain controller must not take the backup estate with it.
  • Patch the backup software itself. Public reporting documents ransomware operators exploiting credential-disclosure flaws in backup platforms (such as Veeam's CVE-2023-27532) to obtain environment credentials for lateral movement. Keep management ports closed externally and versions supported.
  • Sync is not backup. Cloud drive sync and real-time NAS sync push encrypted files off-site during a ransomware event; use versioned backup with a retention window long enough to matter.
  • Verify backups by actually restoring them. Each quarter, fully restore a different data type and record real RTO/RPO, and run periodic consistency checks on backup files rather than watching job success rates.
  • Monitor the backup estate for anomalies. Alert on jobs being disabled, retention being shortened, repositories being deleted, restore points disappearing en masse, and unusual backup-account logons — these are often the last warning before encryption begins.
  • Keep a cold-start package for critical systems. Store system images, configuration documentation, licences, keys and certificates offline, so recovery is not blocked on rebuilding the environment.

Emergency response

Data already encrypted? Stop and let an engineer look first

We do not pay ransoms and we do not negotiate with attackers. Engineers run a free assessment first, then propose a recovery plan and a firm quote.

Related questions

Related industries

FAQ

Frequently asked questions

  • Our backup files are encrypted — are they still worth anything?

    Very possibly. Backup files are usually enormous, and many families encrypt only the header of large files or work intermittently, so complete backup data blocks often survive in the untouched regions. The approach is segmented entropy analysis to establish coverage, then rebuilding the header or extracting content directly from unencrypted segments. The result may be a few usable restore points, or a handful of extractable VMs and databases, rather than an intact chain — but when every other path has failed, that is frequently decisive. So do not delete encrypted backup files.

  • We had off-site sync — why are the off-site files encrypted too?

    Because sync is not backup. A sync job's purpose is to keep both ends identical: once local files are encrypted, it pushes the "new version" — the encrypted files — off-site and overwrites the good copy. Cloud drive sync, real-time NAS sync and folder mirroring all share this flaw. The fix is versioned backup: retain enough history with a sensible retention window, or move to immutable backups and object lock. This is the central item in rebuilding the backup architecture after recovery.

  • How were the attackers able to delete our backups?

    Usually one or a combination of three causes. Reachable permissions: the backup target was a share the production server could write to, so the encryptor simply followed those permissions. Shared credentials: the backup server was domain-joined and ran jobs as a domain administrator, so a compromised domain controller took the backup estate too. The backup software itself was exploited: public reporting documents ransomware operators using credential-disclosure flaws in backup platforms to extract environment credentials for lateral movement. The three remediations map directly: make the target unwritable from production, give the backup estate its own credentials and network, and keep the backup software patched and supported.

  • Can we still count on shadow copies?

    Not as a primary reliance. Deleting shadow copies is standard behaviour for virtually every major family, usually completed before encryption starts, so an empty Previous Versions list is the norm rather than the exception. Worth noting: the deletion itself is a valuable alerting point — if EDR blocks and alerts at that step, there is often a ten-minute window to intervene. So the right approach is to enable EDR shadow-copy and tamper protection, and to build real recovery capability on offline and immutable copies.

  • When rebuilding backup architecture, what matters most?

    One rule must hold: at least one copy the attacker cannot reach at the moment of the attack. There are several ways to achieve it — offline media (tape, external drives detached after use), immutable storage with object lock or WORM retention, or a backup system with independent credentials outside the production domain. Everything else — multiple versions, off-site copies, encrypted transport, drills — matters too, but if this one does not hold, the rest can fail together in a ransomware scenario. Beyond that, backups must have been validated by a real restore, or they are just an untested assumption.

Updated