Backup immutability means a backup copy cannot be changed, encrypted or deleted until a retention date you set has passed, even if an attacker holds an administrator password. The lock is enforced by the storage itself, not by the permissions of whoever is logged in. That is why it is now the baseline for ransomware recovery. It only works when it is configured correctly, kept apart from your everyday admin accounts, and proven with real restores. Those are three jobs that are easier to hand to one accountable IT and security provider than to split across several vendors.
TL;DR
- An immutable backup is write-once for a defined retention period. Stolen credentials cannot shorten that period or delete the copy.
- Immutability and air gapping answer different questions. Immutability stops changes to data you can reach. An air gap stops the data being reachable at all. Strong plans use both.
- Set it up with a locked retention policy, a separate identity for backup administration, encryption with protected keys, and a second isolated copy.
- An immutable copy can still be a copy of infected data, so restore tests and monitoring matter as much as the lock.
- Healthcare organizations should treat tested, encrypted, recoverable backups as a HIPAA Security Rule priority.
Table of Contents
- What backup immutability actually is
- Why attackers go after backups first
- Immutable versus air-gapped backups
- How to make backups immutable, step by step
- Backup encryption and key management
- What immutability does not do
- Healthcare: backups and the HIPAA Security Rule
- Perspective
- Put locked backups in place with Nexus
- FAQ
- Sources
What backup immutability actually is
A normal backup repository is a set of files that your backup software, and anyone with the right rights, can modify or remove. That is convenient for housekeeping. It is also exactly what an attacker needs.
An immutable backup uses write-once, read-many (WORM) storage, often called object lock in cloud storage. Each locked copy carries a retain-until date, and the storage rejects any delete or overwrite request before that date. Because the rule sits in the storage layer, it does not depend on the permissions an intruder could change with stolen credentials. The retention window can run from days to years, depending on your regulatory and business needs.
Two practical points follow:
- Immutability is time-boxed. Once the retain-until date passes, the copy can be deleted again. Your retention policy decides how far back you can recover.
- Not every "immutable" setting is equally strong. In a governance-style mode, privileged roles can still override the lock. In a compliance or locked mode, nobody can. Microsoft's documentation states that in the Locked state of Azure Backup immutability, no user can delete the backup data or the vault, and it recommends that mode for production data to protect against compromised credentials and insider threats.
Why attackers go after backups first
Modern ransomware crews know that a working backup removes their leverage. So they look for the backup server, the backup console and the repository credentials before they trigger encryption. CISA's August 2026 #StopRansomware advisory on Gunra ransomware urges organizations to keep offline, immutable backups in a physically separate, segmented location and to test them, so that recovery does not depend on paying a ransom.
The mechanism is simple. Ransomware encrypts files by issuing write operations. If the storage refuses those writes on locked objects, the attacker can encrypt your servers but not the recovery points. You rebuild from the locked copy instead of negotiating.
For a small business this is the difference between a bad week and a closure. For a healthcare provider it is the difference between restoring patient systems and diverting patients. For a startup running everything in the cloud, it protects against a stolen cloud admin key wiping production and backups in one go.
Immutable versus air-gapped backups
People use the two terms as if they were the same. They are not, and the difference decides what you build.
| Immutable backup | Air-gapped backup | |
|---|---|---|
| What it prevents | Changing or deleting data within the retention period | Reaching the data at all from the network |
| Where it lives | Usually online, in locked storage | Offline media or a logically or physically isolated vault |
| Recovery speed | Fast, because it is immediately accessible | Slower, especially when media is stored offsite |
| Weak spot | Still reachable, so it can be targeted through the backup account or provider control plane | Can be mutable if it is not also locked, and costs more to operate |
A copy can be air-gapped but changeable, or immutable but network-reachable. The strongest design has both properties. A common approach is to keep a locked copy online for fast daily recovery, then hold a second, isolated copy as the last resort. This extends the classic 3-2-1 rule (three copies, two media types, one offsite) with an extra offline or immutable copy. We walk through that in our guide to the 3-2-1 backup rule with an offline copy.
Which control you lean on depends on the failure you are planning for. Accidental deletion and ransomware on a single workstation are handled well by locked storage. A compromised domain administrator or a hijacked backup platform account calls for a copy that sits in a different identity domain and is isolated from the network.
How to make backups immutable, step by step
This is the routine, with where it usually goes wrong at each step.
-
List what must be recoverable, and how fast. Start with a business impact analysis: which systems, how much data loss is tolerable, how long you can be down. Without this, retention and cost decisions are guesses. Our disaster recovery planning guide shows how to start. Where it goes wrong: treating every system as equally critical, which inflates storage cost, or treating none as critical, which leaves gaps.
-
Choose the locked storage target. Options include object storage with object lock, a vault feature in your cloud platform, or a hardened repository on premises. Cloud providers document their own approaches. AWS recommends, for logically air-gapped vaults, Vault Lock in compliance mode, minimum and maximum retention periods, a dedicated account for the vault and multi-party approval for recovery. Where it goes wrong: leaving the lock in a mode that a privileged role can undo.
-
Set retention deliberately. Microsoft's architecture guidance gives a common baseline for critical workloads: 7 to 35 days of short-term retention plus at least 14 to 30 days of immutable retention. Attackers often sit inside a network for days before striking, so a retention window shorter than the intruder's dwell time can leave you with only compromised restore points. Where it goes wrong: a seven-day window on a threat that was planted three weeks earlier.
-
Lock it only after you have validated the policy. The same Microsoft guidance advises locking immutability after policy validation so that privileged admins cannot disable it during an attack. A locked policy with a mistake in it is hard to correct, so test on a non-critical workload first. Locking in Azure also blocks subscription cancellation until the resources are deleted, which is worth knowing before you lock anything you plan to retire.
-
Separate backup administration from production administration. Microsoft recommends separate privileged groups, privileged identity management, conditional access and break-glass accounts for backup. If the same account runs your servers and your backups, one phished password reaches both. Where it goes wrong: a single shared admin identity, or no multi-factor authentication on the backup console.
-
Keep a second, independent copy. Microsoft suggests two independent immutable copies in separate subscriptions and regions. For the last-resort copy, add isolation: a separate identity domain and, where the risk justifies it, a physical air gap.
-
Restore on a schedule. A successful backup job does not prove you can meet your recovery time. Microsoft advises running restore drills from both primary and secondary vaults. Our guidance on how often each system tier needs a recovery test covers the cadence.
If you run an online store, the same logic applies to platform-level data. Choosing backup plugins for BigCommerce is one piece of the picture, and retention and restore testing still decide whether the copy is usable.
Backup encryption and key management
Immutability protects a backup from being altered. Encryption protects it from being read. You want both, because attackers increasingly steal data before they encrypt it, and a locked but readable copy does nothing against theft.
The usual design is envelope encryption: a data key encrypts the backup, and a master key held in a key management service or hardware security module encrypts the data key. This lets you rotate the master key without re-encrypting all the data. Backups are encrypted at rest and in transit, and the keys are managed separately from the backups themselves.
NIST's OT Backup Quick Start Guide points to mechanisms for backup integrity and availability such as hashing, encryption and write-once media. It also advises storing file hashes of backup content so you can verify them before restoring, and running recurring restore tests on non-production systems.
Key handling is where most encrypted-backup plans quietly fail:
- Keys do not belong in scripts or configuration files.
- Whoever administers backups should not automatically have access to the keys.
- A key that an attacker can delete or disable defeats the whole design. This is why AWS recommends AWS-owned keys for its logically air-gapped vaults: the provider manages the key, so an intruder with account access cannot remove it.
- Document who can use which key, how it rotates and how you recover it. Our guide to audit-ready encryption key management sets this out.
What immutability does not do
Knowing the limits is part of using the control well.
- It does not check that data is clean. A locked snapshot of an already infected database is still infected. Pair immutability with 24/7 monitoring so that you spot an intrusion early and know which restore point predates it.
- It is not a substitute for isolation. Locked storage that stays online can still be reached through the backup account or the provider's control plane. That is the case for a second, isolated copy.
- It is not a recovery plan. The lock keeps the data. You still need tested runbooks, ordered restore priorities and people who know who calls whom. A ransomware tabletop exercise is a cheap way to find the gaps, and our hour-one ransomware response guide covers the first hours.
- It costs storage. You pay to keep data you can no longer delete, so retention should match the business need, not the maximum.
Healthcare: backups and the HIPAA Security Rule
For US healthcare providers and their business associates, backups of electronic protected health information fall under the HIPAA Security Rule. In January 2025, HHS published a proposed update. According to law firm ArentFox Schiff, it would make encryption of ePHI mandatory rather than addressable, require retrievable exact copies with verified restoration, and allow a transition period of up to 14 months for existing business associate agreements after a final rule. The same firm notes that the outcome of the proposal is uncertain, but expects the Office for Civil Rights to keep its enforcement focus on Security Rule compliance either way.

The practical reading: build encrypted, immutable, regularly restored backups now, and keep evidence of the restore tests. Our guide on HIPAA-compliant backups and proving your restores goes into the documentation side. Where state law or your counsel's advice sets retention for medical records, follow that, and take specific HIPAA questions to your compliance officer or healthcare attorney.
Perspective
Most of the ransomware recoveries that go badly share one pattern: the backups lived in the same trust boundary as everything else. One credential, one console, one vendor for storage and another for monitoring, and nobody owning the whole chain.
That is why we consolidate. When backup and disaster recovery, data encryption, 24/7 threat detection and compliance assessments sit under one provider with one SLA, the question "who is responsible when the restore fails?" has a single answer. Our approach is designed so that the person who watches for the intrusion is working from the same plan as the person who restores the data.
For a small business, that means locked backups without hiring a storage specialist. For a healthcare provider, it means backup evidence that lines up with HIPAA assessments. For a startup in AWS, Azure or Google Cloud, it means retention and vault design that grows with the environment. Our pricing is flat-rate and monthly, so protection does not arrive with a surprise invoice.
Immutability is the right control. We make sure it is configured, separated, monitored and tested.
Nick - Sr. Executive
Put locked backups in place with Nexus
Nexus provides backup and disaster recovery, data encryption, business continuity planning, 24/7 threat detection and compliance assessments for SOC 2, HIPAA, PCI-DSS and ISO 27001, under one unified SLA and a single point of accountability. If you want immutable backups designed, separated from your admin accounts and restore-tested as part of one managed service, see our IT and cybersecurity services. If you are still sizing the risk, a ransomware readiness assessment shows where your recovery plan stands.

FAQ
Can immutable backups be deleted?
Not before the retain-until date, and not by any user in a compliance or locked mode, including administrators. After the retention period ends, the copy can be removed under your normal policy. In a governance-style mode, privileged roles can still override the lock, so choose the stricter mode for production data.
What is the difference between offline and immutable backups?
An offline backup is disconnected from the network, so an attacker cannot reach it. An immutable backup may stay online, but it cannot be modified or deleted during its retention period. A copy can be one, the other or both, and the strongest plans combine them.
Do immutable backups protect against ransomware?
They protect the recovery points from being encrypted or deleted, which is what lets you restore without paying. They do not tell you whether the data inside was already infected, so combine them with monitoring and regular restore tests. Separate backup credentials from production ones as well.
Should backups be encrypted if they are already immutable?
Yes. Immutability stops changes, and encryption stops someone reading a stolen copy. Keep the encryption keys apart from the backup administrators' everyday access, and document how keys are rotated and recovered.
How long should immutable retention be?
It depends on your recovery needs and any rules that apply to your data. Microsoft's guidance gives a baseline for critical workloads of 7 to 35 days of short-term retention plus at least 14 to 30 days of immutable retention. Regulated organizations often need longer, so set it from your business impact analysis.
Sources
- CISA #StopRansomware: Gunra Ransomware (AA26-222A)
- Microsoft: Design a ransomware-resilient backup architecture
- Microsoft: Azure Backup security best practices for data protection
- AWS re:Post: Building isolated, immutable backups for ransomware recovery
- NIST SP 1339: OT Backup Quick Start Guide
- ArentFox Schiff: Proposed changes to the HIPAA Security Rule
