in one lineBackup deletion is when someone holding an admin account deletes your backups, shortens how long they are kept or turns off their protections, so that nothing is left to restore.
the numbers
- 94%of ransomware attacks went after the backupsSophos, 2024
- 57%of attackers' attempts to compromise backups succeeded (self-reported)Sophos, The impact of compromised backups on ransomware outcomes, 2024
- 67%of organisations with compromised backups paid the ransom, against 36% whose backups held (self-reported)Sophos, The impact of compromised backups on ransomware outcomes, 2024
1what it is
Backups exist for the day production is gone, and ransomware crews plan for the same day. Once an intruder holds an admin account, the backups are what stand between them and a payment, so removing them is part of the job. MITRE ATT&CK, the public catalogue of attacker techniques, files this under Inhibit System Recovery (T1490): deleting shadow copies and backup catalogues on servers and, in cloud environments, switching off versioning and backup policies and deleting snapshots.
CISA's #StopRansomware Guide, revised with the FBI, NSA and MS-ISAC in October 2023, says most ransomware actors try to find accessible backups and then delete or encrypt them, so that restoring is impossible without paying. It adds that they collect credentials inside the network and use them to sign in to the backup system itself.
Deleting recovery points is the loud version. The quiet one is a settings change: retention cut from months to a day, soft delete switched off, an immutability setting removed. The jobs can keep running while the older copies expire on schedule. For the retention windows that Microsoft 365 and Google Workspace keep by default, see the SaaS backup topic. This page is about who can change them.
2why it accumulates
One login reaches both. The backup vault sits in the same cloud subscription as production, the backup server is joined to the same domain, or the IT provider keeps one admin account for the servers and the backup console. The backup admin guide walks through each path, and the IT provider guide covers who at the provider holds that login.
Backup rights sit outside the review. The right to delete a recovery point or shorten retention lives in the backup tool's own role list or in the cloud subscription's roles. An access review that looks at the directory and the business apps can miss it entirely. A top-level cloud admin can also gain those rights by inheritance, which is why blast radius is measured per identity across every system it can reach.
The safety catch belongs to the same admin. Many locks can be lifted by whoever holds enough rights. Microsoft Threat Intelligence described this on 27 August 2025 in its report on a group it tracks as Storm-0501. Through a synced, non-human account with the Global Administrator role and no MFA method registered, the group gave itself the Owner role on the victim's Azure subscriptions. It then deleted snapshots, restore points, Recovery Services protection containers and storage accounts, and removed resource locks and immutability policies where earlier deletions had failed. For resources still protected by immutability policies, Microsoft says the group turned to encrypting them instead.
3what it costs you
Sophos, a security vendor, commissioned a survey of 2,974 IT and security staff at organisations hit by ransomware, run in early 2024. In that vendor research, 94% said the attackers tried to compromise their backups, and Sophos counted 57% of those attempts as successful. Organisations whose backups were compromised paid a ransom to recover encrypted data in 67% of cases, against 36% where the backups held, and their median recovery cost was $3 million against $375,000. The numbers are self-reported, and Sophos itself says there are likely multiple reasons behind the cost gap, including weaker defences overall.
At the far end of the scale, Code Spaces closed in 2014 after an attacker in its AWS console deleted production and backups together.
what it looks like when it goes wrong
Three ways this tends to go, each one taken from a real case.
- it happenedCDK Global, 2024howYour main software vendor is hit and goes dark for two weeks.what leavessales, orders and invoicingwho ends up with itcompetitors, as customers go elsewhere
- it happenedCode Spaces, 2014howBackups live where the attacker can delete them too.what leaveseverything, with nothing left to restorewho ends up with itthe business itself, which closes
- it happenedMaze / Allied Universal, 2019howAttackers copy the data before they lock it.what leavesfiles your backups cannot take backwho ends up with itthe public, when it is leaked
4what works for a small company
List the three rights. For each backup location, write down who can delete backups, who can shorten retention or switch off soft delete and locks, and who can grant those rights to someone else. Mark everyone on the list who also holds admin on production or the directory. That overlap is the problem to shrink.
Give the backups their own admins. Separate accounts with MFA for the backup console, kept off the production domain admin. If a provider runs the backups, CISA's guide asks that the provider follows the same practices, written into the contract.
Keep one copy no admin can remove early. AWS documentation says that in S3 Object Lock compliance mode a locked object version cannot be deleted by any user, including the root user, and its retention cannot be shortened. Governance mode can be overridden by anyone with the bypass permission. In Azure Backup, Microsoft Learn describes multi-user authorization: a Resource Guard owned by a different user must approve critical operations such as disabling soft delete, reducing retention or disabling immutability, and it can sit in another subscription or tenant. CISA cautions that immutable storage can be costly to misconfigure, so set the period with care.
Alert on the quiet changes. The UK National Cyber Security Centre's principles for ransomware-resistant cloud backups ask for alerts on mass deletion requests, backups stopping and changes to global retention periods, raised whether or not the attempt succeeds. Someone has to read them.
Plan the way in for recovery day. NCSC also asks that customers can reach the backup service even when every corporate system is down. The break-glass guide covers how to keep such an account separate, tested and watched.
If the one admin login with the most rights were stolen tonight, which of your backups would still be there on Monday?
on the record
your turn
what you can do about it
this week
List every account, group and provider login that can delete your backups, shorten their retention or turn off soft delete.
Then mark the ones that also hold admin on production or the directory. Each of those is one stolen login away from leaving you nothing to restore.
go deeper
if you're not sure where to start
Run the offboarding checklist on one leaver. It covers every account and token to close when someone leaves.
open the checklist →
