A 2016 ransomware alert called offline backups ideal; its May revision said connected ones could be affected
DHS’s March 31, 2016 ransomware alert said backups should ideally be stored offline. A May 6 revision warned that network-connected backups can also be affected. Where a backup sits, and who can reach it, became the control that decides recovery.
Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.
Two-minute summaryThe dek and the first sentence of each section, quoted from the article
DHS’s March 31, 2016 ransomware alert said backups should ideally be stored offline. A May 6 revision warned that network-connected backups can also be affected. Where a backup sits, and who can reach it, became the control that decides recovery.
- The week The alert opened with a description of the moment.
- Recovery: where the backup is kept The March text said backup data should ideally be kept on a separate device and stored offline. [1] Ideally is a soft word.
- Execution: what is allowed to run The alert recommended application whitelisting, which it described as allowing only specified programs to run while blocking all others, including malicious software. [1] It called it one of the best security strategies. [1] That is the alert’s assessment, and we report it as such.
- Exposure: how ransomware arrived The alert named phishing emails with malicious attachments as a common route. [1] It said that, for enterprises, it may be best to block email messages with attachments from suspicious sources. [1] That was a filtering decision for the email platform, not a user habit.
- Response: the payment decision and reporting The alert treated payment as a practical gamble.
- How the alert reaches different organizations The alert named healthcare facilities and hospitals as victims of both variants it highlighted. [1] For a hospital, isolated backups protect clinical and business data alike.
- Objections to the argument, in order They were, and the March alert listed them first. [1] Our argument concerns placement.
- Checks a technology leader could run For each backup of critical data, record where it is stored, which networks can reach it and which credentials can change or delete it.
- With hindsight US-CERT kept revising the alert after March.
The week
On March 31, 2016, US-CERT, part of the Department of Homeland Security, released Alert TA16-091A, titled Ransomware and Recent Variants. [1] We read the alert as first published, in an Internet Archive copy of the US-CERT page captured on April 1, 2016. [1]
The alert said DHS was releasing it in collaboration with CCIRC, the Canadian government’s cyber incident response body. [1] Its stated purpose was to give further information on ransomware: its main characteristics, its prevalence, variants that may be proliferating, and how users can prevent and mitigate it. [1] It listed the systems affected as networked systems. [1]
The alert opened with a description of the moment. It said that in early 2016, destructive ransomware variants such as Locky and Samas were observed infecting computers of individuals and businesses, including healthcare facilities and hospitals worldwide. [1]
It defined ransomware as malicious software that infects a computer and restricts users’ access until a ransom is paid. [1] It said the ransom demanded from individuals varies greatly but is frequently $200 to $400, paid in virtual currency such as Bitcoin. [1]
The alert described how infections spread. It said ransomware is often spread through phishing emails with malicious attachments or through drive-by downloading from infected websites. [1] It said newer methods had been observed, including exploiting vulnerable web servers as an entry point into an organization’s network. [1]
It traced the growth of more destructive variants. The alert said that in 2013, variants including Xorist, CryptorBit and CryptoLocker were introduced. [1] It said some variants encrypt the files on the infected device and also the contents of shared or networked drives, which is why they were considered destructive. [1]
It then described the two variants of early 2016. Locky, it said, was observed infecting computers of healthcare facilities and hospitals in the United States, New Zealand and Germany. [1] It spread through spam emails with malicious Microsoft Office documents or compressed attachments, using macros or JavaScript to download the ransomware. [1]
Samas, the alert said, was used to compromise the networks of healthcare facilities in 2016. [1] Unlike Locky, it propagated through vulnerable web servers. After a server was compromised, uploaded files were used to infect the organization’s network. [1]
The alert linked ransomware to other malware. It said systems infected with ransomware are also often infected with other malware, and described CryptoLocker infections delivered alongside GameOver Zeus, a banking Trojan. [1] It noted an international law enforcement operation in June 2014 that weakened the infrastructure of both. [1]
On impact, the alert listed temporary or permanent loss of information, disruption of operations, financial losses to restore systems and files, and potential harm to reputation. [1] It said paying a ransom does not guarantee encrypted files will be released, and decrypting files does not mean the infection has been removed. [1]
The alert closed with preventive measures. The first was to employ a data backup and recovery plan for all critical information, and to perform and test regular backups. [1] The original text added: “Ideally, this data should be kept on a separate device, and backups should be stored offline.” [1]
The other measures were to use application whitelisting, keep operating systems and software patched, and maintain up-to-date antivirus software. [1] The alert also advised restricting users’ ability to install and run software under the principle of least privilege, avoiding enabling macros from email attachments, and following safe browsing practices. [1]
It said individuals or organizations are discouraged from paying the ransom, because payment does not guarantee files will be released. [1] It asked that instances of fraud be reported to the FBI’s Internet Crime Complaint Center. [1]
Alert TA16-091A was a broad primer on ransomware, released while hospitals were among its named victims. [1] Its first recommendation was to keep backups, test them, and ideally store them offline. [1] US-CERT revised that sentence on May 6, 2016. [2]
The revised text adds a warning. It reads: “Note that network-connected backups can also be affected by ransomware; critical backups should be isolated from the network for optimum protection.” [2] The page’s revision history records the change as clarified guidance on offline backups. [2] In the two versions we compared, the other differences were a dropped heading label and an added link. [1] [2]
Our argument is that the revision moved the backup recommendation from a matter of having copies to a matter of where they sit and who can reach them. A backup that the infected network can reach is part of what ransomware can encrypt. Later federal guidance built on the same point, and Treasury’s sanctions office counted offline backups among the steps it would treat as mitigating. [3] [4]
We take the alert’s preventive measures in four parts, grouped by what each protects against:
- Recovery: backups, and their position relative to the network.
- Execution: controls on what software can run and with what rights.
- Exposure: the entry points the alert named: email, browsing and web servers.
- Response: the payment decision and reporting.
The parts do not overlap. Exposure concerns how ransomware arrives, execution whether it can run, recovery whether the organization can restore without the attacker, and response what the organization decides once it is hit. After them we look at how the alert reaches different organizations, take the objections in order, and end with checks a technology leader could run.
Recovery: where the backup is kept
From ideal to isolated
The March text said backup data should ideally be kept on a separate device and stored offline. [1] Ideally is a soft word. It allows a reader to treat offline storage as an aspiration and network-attached storage as good enough.
The May text closed that reading. It said network-connected backups can also be affected and that critical backups should be isolated from the network. [2] The revision did not add a new control. It explained why the existing one failed when placed in the wrong spot.
Why connected backups fail
The alert itself had already described the mechanism. It said some variants encrypt the contents of shared or networked drives as well as local files. [1] A backup written to a network share is, from ransomware’s point of view, another networked drive.
The 2023 #StopRansomware Guide described a second mechanism. It said ransomware actors often hunt for and collect credentials stored in the targeted environment, and use them to try to access backup solutions. [4] It said they also use publicly available exploits to target unpatched backup solutions. [4] An isolated backup defends against the first mechanism; separate credentials and patching defend against the second.
Isolation is an architecture decision
Where a backup is kept is decided when the backup system is designed, not when ransomware arrives. The choices include removable media taken offline and a separate network segment with no route from user systems. They also include a provider holding copies outside the organization’s control plane, and storage that cannot be changed for a set period.
Each choice has costs. Offline media is slower to restore. A separate segment needs its own management. Provider copies depend on the provider’s controls. The 2023 guide said immutable storage should be used with caution, because it does not meet compliance criteria for certain regulations and misconfiguration can impose significant cost. [4]
How far back, how fast
Two numbers describe a backup’s value in a ransomware event. The first is how much recent work would be lost if the organization restored from its newest clean copy. The second is how long the restore would take. Both depend on design choices made long before an incident.
Isolation can push both numbers the wrong way. A copy taken offline once a week loses more recent work than a copy written every hour to a connected share. The design task is to find copies that are both isolated and recent enough, and to state the trade-off openly to the business owners who must live with it.
Those owners should agree the numbers in writing for each critical service. When they do, the backup design has a target. When they do not, the target is set by default, by whatever the backup team could build with the budget it had.
Cloud sync is not a backup
The 2023 guide added a warning that the 2016 alert did not need to make. It said automated cloud backups may not be sufficient, because if local files are encrypted by an attacker, they will be synced to the cloud, possibly overwriting unaffected data. [4]
Synchronization copies changes, including harmful ones. A system that mirrors a file share to cloud storage within minutes will mirror the encryption too. A backup worth the name keeps earlier versions that the attacker cannot reach or overwrite.
Encrypted copies
The 2023 guide asks for offline, encrypted backups. [4] Encryption protects the copy’s confidentiality if the media is lost or the storage is reached. It adds a dependency: the keys must survive the incident and be available to the people restoring.
Key management therefore belongs in the recovery design. If the keys are held only in a system the ransomware can reach, the encrypted backup can become as unusable as the encrypted production data. Keys need their own isolated, tested copy.
Testing the restore
The March alert already said to perform and test regular backups. [1] The 2023 guide said to regularly test the availability and integrity of backups in a disaster recovery scenario. [4] Both point at the same gap: a backup that has not been restored is a hope.
A restore test measures more than whether files come back. It measures how long the restore takes, which systems must be rebuilt first, and whether the restored data is current enough for the business. Those numbers decide whether the organization can refuse a ransom in practice.
How many generations to keep
The alert said systems infected with ransomware are also often infected with other malware. [1] An intruder may therefore have been present before the encryption began. A backup taken after that point may contain the intruder’s tools as well as clean data.
Keeping several generations of isolated backups gives the restore team a choice. They can go back to a point before the likely intrusion for system files, while taking more recent copies of data that can be checked. The retention period is a design decision with a storage cost, and it should be set with that scenario in mind.
Rebuilding as well as restoring
The 2023 guide extended the backup idea from data to systems. It recommended maintaining and regularly updating golden images of critical systems, and keeping backups of infrastructure-as-code templates offline. [4] It recommended storing source code or executables with offline backups, along with escrow and license agreements. [4]
That extension reflects the alert’s own warning that decrypting files does not mean the infection is gone. [1] An organization that restores data onto a compromised system has restored the attacker’s foothold too. Clean images let it rebuild the system before putting the data back.
Backups of the systems that run the restore
A restore depends on more than the data. It depends on the directory that authenticates administrators, the backup software and its catalog, the network services that let systems find each other, and the documentation that says what to restore first. If those are encrypted too, the data copies may be intact and still hard to use.
The 2023 guide’s recovery checklist told organizations to restore data from offline, encrypted backups based on a prioritization of critical services, and to take care not to reinfect clean systems during recovery. [4] Prioritization needs a list, and the list needs to be available when the network is not. Recovery planning should include the systems and records the restore itself needs.
What this means for platform owners
For the owner of a storage or backup platform, the alert’s revision sets three design questions. Can a user’s or administrator’s ordinary credentials delete or change backup copies? Is there a network path from user systems to backup storage? How far back can the organization restore, and how fast?
Those questions are architectural. They are answered by network design, identity design and product configuration. A policy that says to keep backups offline does nothing until someone builds the platform so that it is true.
Execution: what is allowed to run
Application whitelisting
The alert recommended application whitelisting, which it described as allowing only specified programs to run while blocking all others, including malicious software. [1] It called it one of the best security strategies. [1] That is the alert’s assessment, and we report it as such.
Whitelisting changes the question from which programs are bad to which are allowed. For ransomware delivered as a new executable, it can stop the payload from running at all. Its cost is maintenance: every legitimate program and update must be approved, and exceptions accumulate.
The 2023 guide uses the term application allowlisting for this kind of control. [4] The 2016 alert used whitelisting, and we keep its word when describing it.
Least privilege
The alert advised restricting users’ permissions to install and run unwanted software, and applying the principle of least privilege to all systems and services. [1] It said restricting privileges may prevent malware from running or limit its ability to spread. [1]
Spread matters as much as initial infection. Ransomware that runs with a user’s rights can encrypt what that user can reach. Ransomware that obtains administrator rights can reach much more, including, in a poorly designed environment, the backups. Least privilege is therefore part of the backup design as well as the endpoint design.
Macros
The alert advised users to avoid enabling macros from email attachments, explaining that enabling them lets embedded code execute the malware. [1] It said Locky spread through Office documents and compressed attachments using macros or JavaScript. [1]
Advice to users is the weakest form of macro control, because it depends on each person making the right choice each time. An organization can enforce the choice instead, by blocking macros in files from the internet or allowing only signed ones. That moves the decision from the user to the platform owner.
Scripts and documents
The alert described Locky using macros or JavaScript files to download the ransomware. [1] Those are scripts and document features, not standalone programs. An allowlisting policy aimed only at executable files may let them run. Script hosts, document macros and command interpreters need their own rules.
Each rule has a cost in disruption. Finance teams may depend on spreadsheet macros; administrators on scripts. The platform owner should find those uses first, then allow them narrowly, rather than leave a broad opening because some legitimate work needs a small one.
Administrator accounts
The alert’s least privilege advice applied to all systems and services. [1] Accounts with broad administrative rights carry particular weight in a ransomware attack: domain administrators, virtualization administrators and backup administrators. An attacker holding one of them can push ransomware widely or reach the copies meant for recovery.
Separating those roles, so that no single account administers both production systems and backups, limits what one stolen credential can do. Requiring stronger sign-in for them, and logging their use to a place they cannot alter, gives the organization a chance to see misuse before the encryption starts.
Execution controls as architecture
Whitelisting, privilege and macro settings are configured centrally. They belong to the people who run endpoint management, identity and productivity platforms. The alert addressed users and administrators together; in a larger organization, those are different teams with different owners.
For a technology leader, the point is to assign each control to the platform that enforces it, and to measure its coverage. A whitelisting policy that covers half the endpoints protects half the endpoints.
Exposure: how ransomware arrived
The alert named phishing emails with malicious attachments as a common route. [1] It said that, for enterprises, it may be best to block email messages with attachments from suspicious sources. [1] That was a filtering decision for the email platform, not a user habit.
Browsing
The alert named drive-by downloading, in which a user visits an infected website and malware is installed without the user’s knowledge. [1] It advised safe browsing practices and not following unsolicited web links in emails. [1] Its advice to keep software up to date with the latest patches applies to browsers as much as to any other program. [1]
Web servers
The exposure section’s description of Samas raises architectural questions. The alert said Samas propagated through vulnerable web servers, and that after a server was compromised, uploaded files were used to infect the organization’s network. [1]
That route did not need a user to click anything. It needed an internet-facing server with a known weakness and a path from that server to the internal network. Both are design and operations questions: patching the server, and segmenting it from the rest of the environment.
Patching
The alert said vulnerable applications and operating systems are the target of most attacks, and that patching greatly reduces the number of exploitable entry points. [1] Those are the alert’s statements. The operational task behind them is an inventory of internet-facing systems and a patch process that reaches them first.
The 2023 guide later added an initial access vector for internet-facing vulnerabilities in its October 2023 version. [4] The addition echoes the Samas route the 2016 alert described. [1]
Shared and networked drives
The alert said some variants encrypt the contents of shared or networked drives, as well as files on the infected device. [1] That makes file share permissions part of the exposure. A user who can write to many shares can, once infected, damage all of them.
Reviewing share permissions is unglamorous work. It reduces how far a single infection reaches, and it reduces what must be restored afterward. It also overlaps with the backup question: a backup written to a share that many users can modify is exposed in the same way.
Connections from outside parties
The alert named email, websites and web servers as entry points. An organization’s exposure also includes the connections its vendors and partners use to reach its systems. Each is a path that does not pass through the organization’s own users, and each should be inventoried, limited to what it needs and monitored as an entry point in its own right.
Antivirus
The alert advised maintaining up-to-date antivirus software and scanning software downloaded from the internet before running it. [1] Antivirus catches known malware. It depends on the malware having been seen before, and it changes little in the architecture.
Exposure and the healthcare setting
The alert named healthcare facilities and hospitals among the victims of both Locky and Samas. [1] Healthcare networks connect clinical systems, devices, business applications and internet-facing services. In such a network, a compromised web server or a clicked attachment can reach systems that support patient care.
That is why segmentation appears in our exposure part and our recovery part alike. It limits how far an intruder can move after entry, and it keeps backups out of reach. The same network design decision serves both purposes.
Response: the payment decision and reporting
What the alert said
The alert said individuals or organizations are discouraged from paying the ransom, as payment does not guarantee files will be released. [1] It said payment only guarantees the attackers receive the victim’s money and, in some cases, banking information. [1] It asked that fraud be reported to the FBI’s Internet Crime Complaint Center. [1]
The alert treated payment as a practical gamble. It did not discuss legal risk. Later federal guidance added that dimension.
Containment first
Before any payment decision comes containment. The 2023 guide’s response checklist begins by telling organizations to determine which systems were affected and immediately isolate them. [4] It says that if several systems or subnets appear affected, organizations should take the network offline at the switch level. [4]
The same checklist warns that actors may monitor an organization’s communications to see whether they have been detected. [4] It advises isolating systems in a coordinated manner and using out-of-band communication, such as phone calls. [4] An organization whose response plan assumes working email and chat should revisit that assumption.
Sanctions risk
On September 21, 2021, Treasury’s Office of Foreign Assets Control issued an updated advisory on sanctions risks of facilitating ransomware payments. [3] It updated and superseded an advisory of October 1, 2020. [3] It said the U.S. government strongly discourages all private companies and citizens from paying ransom or extortion demands. [3]
The advisory said companies that facilitate ransomware payments on behalf of victims, including financial institutions, cyber insurance firms and incident response firms, may risk violating OFAC regulations. [3] Organizations considering a payment should take legal advice on sanctions exposure before acting.
Backups as a mitigating factor
The 2021 advisory tied the payment question back to the backup question. It said meaningful steps to reduce the risk of extortion by a sanctioned actor, through adopting or improving cybersecurity practices, will be considered a significant mitigating factor in any OFAC enforcement response. [3]
It listed examples: maintaining offline backups of data, developing incident response plans, instituting cybersecurity training, regularly updating antivirus and anti-malware software, and employing authentication protocols. [3] In our view, that list turned the 2016 alert’s backup advice into evidence an organization might one day need to show.
Insurers and responders
The OFAC advisory named cyber insurance firms and companies involved in digital forensics and incident response among those that facilitate ransomware payments on behalf of victims. [3] It said such companies may risk violating OFAC regulations. [3]
For a victim, that means the people it relies on during an incident have their own compliance obligations. An organization should know, before an incident, how its insurer and its response firm handle payment questions and sanctions screening. Those answers belong in the response plan alongside the contact numbers.
Telling staff and customers
The alert did not address communication, but an organization hit by ransomware has to tell staff what to do and customers what to expect. Staff need instructions that do not depend on the encrypted systems: printed lists, phone trees and agreed alternative channels. Customers need accurate, cautious statements that change as the facts become clear.
Reporting
The alert asked for reports to the Internet Crime Complaint Center. [1] The 2021 OFAC advisory said that, for ransomware payments that may have a sanctions nexus, OFAC would consider a company’s self-initiated and complete report of an attack to law enforcement or other relevant agencies, made as soon as possible after discovery, a voluntary self-disclosure and a significant mitigating factor. [3] It named CISA and Treasury’s Office of Cybersecurity and Critical Infrastructure Protection among them. [3]
A cross-sector reporting rule was still a proposal when we checked. CISA’s May 26, 2026 notice said the proposed CIRCIA rulemaking, published April 4, 2024, seeks to implement covered cyber incident and ransom payment reporting requirements for covered entities. [5] Sector rules may already require reports, and organizations should confirm their own obligations with counsel.
The decision owner
The payment decision touches legal, financial, operational and reputational questions at once. It should not be made for the first time during an incident. An organization can decide in advance who has authority, who must be consulted, how sanctions screening will be done and how the board will be informed.
The 2023 guide recommended that the incident response plan and communications plan be reviewed and approved in writing by the chief executive, or equivalent, and that a hard copy and an offline version be available. [4] A plan stored only on the encrypted network is a plan no one can read during the incident it was written for.
How the alert reaches different organizations
A hospital or health system
The alert named healthcare facilities and hospitals as victims of both variants it highlighted. [1] For a hospital, isolated backups protect clinical and business data alike. Downtime procedures, paper records and manual workflows serve as the operational backup while systems are restored. HIPAA obligations around a ransomware event are a matter for counsel.
An ambulatory network or physician group
Smaller practices may rely on a managed service provider or an EHR vendor for backups. Their questions are contractual: where the copies are kept, whether the provider’s administrative access could be used to delete them, and how fast a restore can be done. Those questions should be asked before signing, and again at renewal.
A bank or credit union
Banks and credit unions have regulators that examine business continuity and data backup. The 2016 alert adds a specific test to that work: whether backups are reachable from the production network or by production credentials. A bank that pays a ransom also faces the sanctions questions the OFAC advisory describes, as a payer and, in some cases, as a facilitator.
A county or city government
Local governments run tax, court, public safety and payment systems, and their security staffing may be limited. An isolated backup can be the difference between a recovery and a long outage. Such governments can draw on the 2023 guide, which listed the Multi-State Information Sharing and Analysis Center among its authors.
A broker-dealer or investment adviser
Securities firms have their own record-keeping and continuity rules, which counsel should confirm. The architecture point carries over unchanged: a backup reachable by ordinary credentials is exposed to the same encryption as the systems it protects.
A managed service provider
A provider that runs backups for many clients holds a concentration of recovery capability. If an attacker reaches the provider’s management tools, every client’s backups may be at risk at once. Providers should show clients how their own administrative access is separated and protected.
A state agency
State agencies run benefit, licensing and revenue systems that residents depend on. They may share infrastructure with other agencies through a central IT department. In that setting, the backup question has two owners: the agency that owns the data and the central department that runs the storage. Both should know where the isolated copies are and who can reach them.
A research or life sciences organization
Research organizations hold data that may be hard or impossible to recreate: experimental results, trial records, instrument outputs. For them, a backup may be the only other copy of a body of work. Isolation and restore testing should cover research storage as well as business systems, which may belong to different teams.
Objections to the argument, in order
1. Backups were already standard advice
They were, and the March alert listed them first. [1] Our argument concerns placement. The March text treated offline storage as ideal; the May text said connected backups can be affected and critical backups should be isolated. [1] [2] That change concerns architecture, which the earlier wording left open.
2. Offline backups cannot meet recovery time objectives
Offline media can be slow to restore. Isolation does not require tape in a vault. It requires that the infected network and its credentials cannot reach the copy. Several designs meet that test with faster restores, each with costs, such as those the 2023 guide noted for immutable storage. [4]
3. Cloud backup solves the problem
The 2023 guide said automated cloud backups may not be sufficient, because encrypted local files can sync to the cloud and overwrite unaffected data. [4] Cloud storage can be part of an isolated design. Sync alone is not.
4. Prevention matters more than recovery
Prevention matters, and three of our four parts concern it. The alert’s own impact section said paying does not guarantee release and decrypting does not remove the infection. [1] When prevention fails, recovery from an isolated backup is the route that does not depend on the attacker.
5. Data theft makes backups less relevant
The 2023 guide addressed ransomware and data extortion together, and described actors exfiltrating data and pressuring victims to pay by threatening to release it. [4] Backups do not prevent a leak. They do decide whether the organization can operate without the attacker’s key, which narrows what the attacker can demand.
6. Sometimes payment is the only option
The OFAC advisory recognized that companies face this decision and set out the sanctions risks and mitigating factors. [3] It said the government strongly discourages payment. [3] An organization with tested, isolated backups is less likely to face a situation where payment looks like the only option. That is the practical value of the May 2016 revision.
7. Small organizations cannot afford isolated backups
Isolation has a cost, and small organizations with small technology budgets feel it sharply. The 2023 guide asked organizations to apply its practices to the greatest extent possible, given the availability of resources. [4] Even a modest design, such as a rotated offline copy of the most critical data, meets the May 2016 wording better than a connected share alone. [2]
Checks a technology leader could run
Map backup reachability
For each backup of critical data, record where it is stored, which networks can reach it and which credentials can change or delete it. Flag every copy that ordinary user or domain administrator credentials can reach.
Separate backup credentials
Give backup administration its own accounts, with strong authentication and no reuse of domain credentials. The 2023 guide described actors collecting credentials to access backup solutions. [4]
Patch the backup platform
Treat backup software and appliances as high-value systems in the patch process. The 2023 guide said actors use public exploits against unpatched backup solutions. [4]
Test restores
Restore a critical system from an isolated backup on a schedule, and time it. Compare the time against the business’s needs. Record the result and the gaps.
Keep clean images
Maintain golden images and offline copies of infrastructure templates for critical systems, as the 2023 guide recommended. [4] Test that a system can be rebuilt from them.
Measure execution controls
Report the share of endpoints covered by application allowlisting, the number of users with local administrator rights, and the macro policy in force. Set targets and owners for each.
Segment internet-facing servers
List internet-facing servers and confirm they cannot reach backup storage or user networks without controls. The Samas route the alert described began at such a server. [1]
Prepare the payment decision
Write down who decides on any ransom payment, who must be consulted, how sanctions screening will be done and when law enforcement and regulators will be contacted. Review it with counsel against the OFAC advisory. [3]
Review share permissions
Find the users and service accounts with write access to many file shares, and reduce it where it is not needed. Recheck after each reorganization or system migration. Pay particular attention to shares that hold backups or exports, since a write permission there reaches the recovery copy.
Rehearse with the business
Run a tabletop exercise in which ransomware has reached the backups that the network can see. Ask the business owners which services they need first and what they can do on paper meanwhile. Record their answers in the recovery plan.
Encrypt and protect the keys
Encrypt backup copies, as the 2023 guide advised, and keep the keys in a place the production network cannot reach. [4] Test a restore that uses only the isolated keys.
Watch the backup system
Alert on deletion of backup jobs, changes to retention settings and logins to backup consoles. Route the alerts to a team that does not depend on the same administrative accounts.
Keep evidence of your controls
Record backup design, restore tests and incident plans with dates. The OFAC advisory said meaningful cybersecurity steps, including offline backups, would be a significant mitigating factor in any enforcement response. [3] Evidence of those steps is easier to produce if it is kept as the work is done.
Print the plan
Keep a hard copy and an offline version of the incident response plan and contact lists, as the 2023 guide advised. [4] Check them at each plan review.
With hindsight
US-CERT kept revising the alert after March. Its revision history lists the initial publication on March 31, 2016, clarified guidance on offline backups on May 6, 2016, and an added link to governmental interagency guidance on ransomware on July 11, 2016. [2] The page shows it was last revised on September 29, 2016. [2] As checked on November 16, 2026, CISA hosts it as archived content. [2]
The May 6 wording is the one that lasted. The alert now says network-connected backups can also be affected and critical backups should be isolated from the network for optimum protection. [2] Later federal guidance kept the idea and added detail.
In September 2020, CISA and MS-ISAC released a joint Ransomware Guide. [4] Its successor, the #StopRansomware Guide, carries an October 2023 publication date and lists CISA, MS-ISAC, NSA and the FBI as its authoring organizations. [4] It says to maintain offline, encrypted backups of critical data and regularly test their availability and integrity. [4]
The 2023 guide explained why in terms the 2016 revision implied. It said most ransomware actors attempt to find and then delete or encrypt accessible backups, to make restoration impossible unless the ransom is paid. [4] It warned that automated cloud backups may sync encrypted files and overwrite unaffected data. [4]
The payment question gained a legal dimension. OFAC’s October 1, 2020 advisory, updated and superseded on September 21, 2021, described sanctions risks for those who make or facilitate ransomware payments. [3] It said meaningful cybersecurity steps, including maintaining offline backups, would be a significant mitigating factor in any enforcement response. [3]
The guide also added a caution the 2016 alert did not consider. It said immutable storage should be used with caution, since it does not meet compliance criteria for certain regulations and misconfiguration can impose significant cost. [4]
Cross-sector reporting of ransom payments remained a proposal as checked on November 16, 2026. CISA’s May 26, 2026 Federal Register notice, the newest CISA notice in the CIRCIA rulemaking that we found there, described the April 4, 2024 proposal as seeking to implement incident and ransom payment reporting requirements. [5]
The same notice said CISA did not hold town hall meetings it had announced in February 2026, because of a lapse in DHS appropriations from February 14 to April 30, 2026. [5] It listed new dates of June 15, 16, 17 and 18, 2026. [5] We did not find a final rule.
The 2023 guide’s response checklist also reflects the 2016 revision. Its recovery steps tell organizations to reconnect systems and restore data from offline, encrypted backups based on a prioritization of critical services. [4] The offline copy the March 2016 alert called ideal had become the step the recovery plan assumes. [1] [4]
What held up from 2016 was the alert’s list of controls: backups, whitelisting, patching, least privilege and macro restrictions all reappear in later guidance. [1] [4] What changed was the weight on where backups sit. The May 2016 revision is a small edit to a public alert. In our reading, it records the moment federal guidance said plainly that a reachable backup is part of the attack surface. [2]
Spotted an error? Tell us and we will check it and correct the article where needed.