Attackers can disable the systems needed to restore data even when backup copies survive. Understanding those dependencies is becoming an important part of ransomware preparedness.
Northbridge Services has a backup problem that its monitoring dashboard cannot explain.
The company’s IT team has maintained daily backups of its critical applications. Backup jobs completed successfully, and the latest reports show that recovery points are available.
Then a ransomware incident disrupts several production servers.
The infrastructure team begins the recovery process, expecting to restore the affected virtual machines. Instead, administrators discover that the backup management platform may have been compromised. The virtualization environment is also under investigation, and the privileged accounts normally used for recovery cannot yet be trusted.
The backup data may still exist. The problem is reaching it, restoring it and bringing applications online without giving the attacker another opportunity to interfere.
Northbridge is a fictional organization, but the problem reflects attack patterns documented by Google’s Mandiant security team.
In its M-Trends 2026 report, Mandiant describes ransomware operators targeting identity services, virtualization infrastructure and backup systems to undermine an organization’s ability to recover.
These attacks expose an important distinction. Protecting backup copies and maintaining a usable recovery process are related responsibilities, but they are not the same thing.
How an identity compromise can reach the recovery systems
To understand Northbridge’s problem, consider how its infrastructure normally works.
Employees and administrators authenticate through a central identity service. The virtualization platform runs the company’s application servers. A backup application connects to that platform, manages backup jobs and keeps a catalog of available recovery points.
This arrangement is convenient. Administrators can manage infrastructure through established accounts, and the backup system can communicate with the platforms it protects.
It also creates relationships between systems that may not be obvious during routine backup checks.
Suppose an attacker gains control of a privileged account with access to the infrastructure management environment. Depending on its permissions, that account could provide a route into the backup application, virtualization platform or supporting services.
The attacker does not necessarily need to compromise every application server individually.
Mandiant reports observing threat actors manipulate identity environments, exploit misconfigurations in identity services and target virtualization management systems. Its 2026 findings also describe attackers deleting backup objects from cloud storage.
Those findings do not establish that every privileged account provides these capabilities. The outcome depends on how identity, permissions and management networks are configured.
For Northbridge, however, the possibility changes the investigation. The team can no longer assume that the accounts and tools normally used for restoration are safe simply because they worked before the incident.
The recovery process itself has become part of the security investigation.
The backups survived, but the recovery path did not
Northbridge’s investigators determine that some protected backup copies remain available. That is encouraging, but the company still cannot begin a normal restoration.
A backup platform needs more than stored data to recover an application.
It may need a functioning management service to locate the correct recovery point, credentials to access storage, encryption keys to read protected data and computing infrastructure on which to restore the workload.
Northbridge also needs its applications to return in the correct order. Its business application depends on a database, and both depend on working network, storage and administrative services.
The company therefore faces a chain of dependencies rather than one missing file.
This is where attacks against virtualization infrastructure become particularly consequential.
A hypervisor allows multiple virtual machines to share physical computing resources. Its management systems can control virtual machines, their configuration and, depending on the platform, aspects of their underlying storage.
An attacker with sufficiently privileged access to this layer may be able to disrupt several systems without first compromising each guest operating system.
Mandiant’s 2026 analysis describes attacks against hypervisor datastores and management infrastructure. These techniques can affect multiple workloads and complicate restoration.
For Northbridge, restoring an application into an environment that may still be controlled by an attacker would introduce another risk. The restored workload could be modified, disrupted or compromised again.
Protected backups help reduce that risk, but they solve only part of the problem.
An immutable backup is designed to resist modification or deletion during its configured retention period. Offline copies offer another protection because they are disconnected from systems that an attacker can reach.
Neither measure automatically provides the credentials, encryption keys, trusted computing environment or application configuration required for recovery.
For example, an immutable database backup may remain intact while the encryption key needed to read it is inaccessible. A complete virtual-machine backup may also be unusable until administrators establish a trusted place to restore it.
The technical lesson is not that immutable backups are ineffective. They are valuable safeguards against one important class of attack.
The limitation is that protecting stored copies does not automatically protect every system required to use them.
Rebuilding from a trusted environment
Northbridge now has to make a difficult decision.
It could attempt to restore its applications using the existing management infrastructure, but investigators have not established whether that environment is trustworthy.
The safer approach in this fictional incident is to establish a clean recovery path before reconnecting restored workloads to production.
The team begins by identifying backup copies that are believed to predate the compromise and confirming that those copies can be accessed through protected credentials.
It also verifies the availability of required encryption keys, system configurations and recovery documentation.
Next, the infrastructure team prepares a trusted environment for restoration. Depending on Northbridge’s architecture, this could involve separate virtualization resources, isolated cloud infrastructure or a recovery environment with restricted management access.
The purpose of isolation is to prevent the compromised production environment from controlling the recovery process.
Accessible text alternative for this figure
The figure shows how a recovery can proceed when the production administration systems are not trusted. It has three zones, from top to bottom, and five numbered steps that follow the article.
Production, trust not established: an identity service, virtualization management and backup administration. Dashed amber lines mark paths that an attacker with a privileged account may be able to use. The normal restore route from backup administration to the protected copies is marked as not trusted yet.
Protected backup copies: stored copies that resist changes during retention but are not enough alone. Step 1: pick a copy believed to predate the compromise. Step 2: confirm keys and protected credentials. Recovery also needs keys, credentials and a trusted place to restore.
One controlled path leads from the protected copies to the isolated recovery environment. No management path leads from production into it.
Isolated recovery environment, with separate and restricted management access. Step 3: prepare the trusted environment. Step 4: restore supporting services, then the database, then the application. Step 5: check for persistence before reconnecting to the production network.
The scenario is fictional. The figure is an illustration, not a vendor architecture or a real deployment.
Once the team has established trusted administrative access, it restores the necessary supporting services before bringing back the business application. Database and application dependencies are checked, and restored systems are examined for evidence of persistence or reinfection.
Only after these checks does Northbridge consider reconnecting services to the normal production network.
This is a simplified scenario. A real incident would require decisions based on forensic findings, application requirements, available infrastructure and the organization’s approved incident-response procedures.
It also shows why recovery architecture involves trade-offs.
An entirely separate infrastructure environment can provide stronger isolation, but it can be expensive to maintain. Running another complete identity and virtualization environment also creates operational responsibilities of its own.
Not every business needs that arrangement.
A smaller organization using managed services may be able to achieve appropriate separation through provider-supported recovery capabilities, protected administrative access and independently accessible backups.
A larger organization with critical applications, extensive virtualization or long recovery dependencies may need more dedicated infrastructure.
The correct design depends on what must remain operational or recoverable after a particular failure. The objective is to remove dangerous shared dependencies without creating unnecessary complexity.
What would prove that recovery actually works?
Northbridge eventually restores its fictional business application. The team verifies that users can authenticate, the application can reach its database and the expected business functions operate correctly.
That result is more meaningful than the successful backup jobs shown before the incident.
But a company should not need to experience ransomware before learning whether its recovery plan works.
Many backup tests restore a file, a database or an individual virtual machine while the surrounding infrastructure remains healthy. These tests are useful because they verify important parts of the backup process.
They do not necessarily demonstrate recovery when the identity service, backup administration platform or virtualization environment is unavailable.
A stronger exercise would simulate those conditions.
Northbridge could begin by assuming its normal privileged accounts cannot be used. The team would then attempt to retrieve a protected recovery point, access the required keys, prepare a trusted restoration environment and recover a complete application with its dependencies.
The exercise should also establish how long restoration takes, which steps require manual intervention and whether the restored service meets the business’s recovery requirements.
This approach is consistent with NIST’s Guide for Cybersecurity Event Recovery, which emphasizes realistic recovery exercises, clear responsibilities and learning from test results.
The CISA StopRansomware Guide similarly recommends offline protection for critical backups and regular testing of their availability and integrity.
For Northbridge, the exercise reveals whether the recovery plan depends on systems that may be unavailable during the incident it is supposed to address.
For another organization, the critical dependencies may be different. The underlying question remains practical: which business service must be restored, what does that restoration require, and can those requirements still be met after the assumed compromise?
An organization can begin answering that question without launching an entirely new infrastructure program. It can choose one important application, map its recovery dependencies and test restoration under a realistic failure condition.
The results will show whether existing protections are sufficient or where a targeted improvement is necessary.
Modern ransomware preparedness still requires reliable backups. Mandiant’s findings do not change that foundation.
They show why the surrounding recovery systems deserve equal attention.
A recovery plan becomes credible when an organization can demonstrate that it can restore the business services it depends on, even when parts of its normal operating environment are no longer trustworthy.
References and further reading
- M-Trends 2026 Executive Edition, Google Cloud / Mandiant. Findings from 2025 incident investigations, including recovery-denial tactics involving identity, backup and virtualization infrastructure.
- M-Trends 2026: Data, Insights, and Strategies From the Frontlines, Google Cloud. Additional details on attackers targeting critical management and recovery infrastructure.
- StopRansomware Guide, CISA and partner agencies, October 2023. Guidance on protecting backup data and testing recovery.
- Guide for Cybersecurity Event Recovery, SP 800-184, NIST, December 2016. Guidance on recovery planning, dependency mapping and realistic exercises.
Source review date: November 16, 2026.
