Your Backup Is Fine, But Your Environment Isn't Coming Back

Cybersecurity Awareness Month is over. The awareness posts have stopped, the webinars have been archived, and many organizations have gone back to whatever they were doing on September 30. So here’s the reminder that outlasts the month: it’s not just your backup.

Ask a CISO how confident they are in their data recovery, and most will say some version of “problem solved.” Ask the same person how confident they are that they can rebuild the actual cloud environment where their data runs, and the answer changes. The pause is usually the most honest part of the conversation.

That gap has a name now. Gartner placed Cloud Application Infrastructure Recovery Solutions (CAIRS) in its 2025 Hype Cycle for Backup and Data Protection Technologies. The category has been forming for a while. Gartner covered adjacent ground in 2023 under Cloud Infrastructure Recovery Assurance Software (CIRAS) in that year’s report, Storage and Data Protection Hype Cycle.

The market is still early, and analysts expect adoption to grow as more teams learn that backup protects the data but not the networks, permissions, and configuration around it. To understand what’s at stake, it helps to look at what a cloud application is made of.

The recovery gap between data and environment

A modern cloud application is a web of dependencies: virtual networks, subnets, security groups, route tables, load balancers, IAM roles, managed databases, container services, and the configuration that ties it all together. Backup tools were built to protect the data inside that web. Capturing the web itself is newer, and fewer tools do it.

That distinction can be easy to ignore until the day it matters completely. A destructive insider with console access can tear down a virtual network in minutes. A bad deployment can cascade through a dozen dependent services before anyone notices. An attacker with stolen admin credentials can rewrite identity policies, security groups, and routes without touching a single file. In every one of those scenarios, the data can be perfectly intact, and the environment it depends on can be gone.

What happens next is usually a spreadsheet, a Slack thread, and someone’s institutional memory of how the network used to be wired. That’s not a recovery plan. It’s an improvisation performed under pressure, during an incident. And the incidents keep arriving faster. CrowdStrike’s 2025 Threat Hunting Report counted more cloud intrusions in the first half of 2025 than in all of 2024, a 136% increase.

Immutable is not the same as clean

Rebuilding a cloud environment means restoring it from something: the data, the configuration, and the definitions that describe how it all connects. Whatever you restore from becomes your new production, so the copy has to be right.

Immutability guarantees the copy can’t be tampered with. An object locked under a compliance policy can’t be altered or deleted before its retention period expires, no matter who asks or what credentials they have. That’s an extraordinarily useful property, and it’s the reason modern object storage has become the backbone of ransomware defense.

Cleanliness is a different guarantee, and it’s about content. It answers the question: Is what’s stored actually safe to restore from? An immutable copy of an already-encrypted file is still encrypted. The lock didn’t fail. It did exactly its job, which was to preserve the object precisely as it was written, including whatever damage was already done to it.

The organizations building genuine cloud recovery capability right now have converged on the same fix: validating a recovery point before trusting it, rather than assuming the most recent copy is automatically the safest one. That validation work (threat scanning, drift detection, configuration comparison) is a distinct discipline from storing the copy in the first place. Teams that assume storage covers validation learn otherwise mid-incident.

Locked doesn’t mean hidden

An attacker with stolen admin credentials can’t alter or delete a locked object, but they can still list your buckets, read what’s inside, and copy it. That matters more than it used to. CrowdStrike found that 81 percent of hands-on-keyboard intrusions during its 2025 reporting period involved no malware. The attacker is using legitimate tools and valid credentials, which means they appear to be an administrator to your storage layer.

Covert Copy™ is built for that case. It creates a locked, hidden copy of the buckets you select, logically air-gapped and invisible from within the account, and accessing it requires Multi-User Authorization. An attacker holding valid console credentials can’t encrypt what they can’t find. Better yet, it’s included with Wasabi Hot Cloud Storage at no additional cost.

The recovery you never rehearse is the recovery that fails

Everything above assumes the recovery works. Testing is how you find out, and a real test covers more than restoring files. It rebuilds the environment (network, identity, and configuration), restores production-scale data into it, and repeats often enough that the team has it in muscle memory before the day it counts.

Most teams don’t test that often. In Forrester’s 2025 survey with the Disaster Recovery Journal, only 40% of respondents regularly failed over to their recovery site at least once a year, and 27% had no plans to. Forrester also notes that many organizations test one component at a time, which doesn’t reproduce a real failure.

Time and staffing are common reasons teams skip tests, but cost is the one a storage vendor can change.  On infrastructure where every gigabyte pulled out for a test carries a transfer fee, rehearsal becomes something that gets scheduled quarterly if at all, reviewed by finance, and dropped when budgets tighten. The pricing model penalizes exactly the behavior that produces resilience.

If the cost of testing a recovery scales with the number of times you test, you’ll test it less than you should.

When you remove egress fees from the equation, you are no longer penalized for testing frequently. Wasabi doesn’t charge egress or data transfer fees. A fair use policy applies and is spelled out in the terms at sign-up. Cost is no longer a reason to skip the test.

What storage does and doesn’t do in recovery

Object storage doesn't discover your cloud resources. It doesn't map dependencies between your compute, your network, and your identity layer. It doesn't detect configuration drift, and it doesn't orchestrate the sequence in which a rebuilt environment comes back online.

Those are real, hard problems, and a growing set of vendors are purpose-built for this challenge and do exactly that job. If your organization is evaluating cloud application recovery, that orchestration layer is where most of the engineering complexity lives, and it deserves its own evaluation on its own terms.

What object storage is responsible for is narrower and foundational: being the place the data lands, in a form nobody can alter after the fact, at a cost that doesn’t punish you for testing whether any of it works.

The questions to ask before the incident

Put these to your own team, or to any vendor pitching you a recovery story:

• If our cloud environment disappeared tomorrow, not just the data but the network, the identity layer, and the configuration, how would we actually rebuild it, and who owns that job?

• How often do we rehearse a full recovery, and why don't we do it more often?

• When we recover, how do we know the point we're recovering to is clean, rather than just the most recent one available?

• Could an attacker holding valid credentials in our primary account even find our recovery copies?

• Does our storage economics reward testing recovery, or discourage it?

None of those questions require knowing the CAIRS acronym. They require being honest about the gap between “our data is backed up” and “we could actually come back from a bad day.”

That’s the reminder: cyber resilience is not just the backup. October gave the topic a month of attention. The gap it points at is a year-round engineering problem, and November is as good a time as any to find out how wide yours is.