Cybersecurity
Ransomware prevention - controls and the order to deploy them
Ten ransomware controls in deployment order: offline backups, patching internet-facing services, MFA, Active Directory, lateral movement and detection per CISA.
Józef SulwińskiPublished Updated 18 min read

Contents
- Deployment order
- Step 1. Record your assets and decide which systems are critical
- Step 2. Keep offline backups and test restoring them regularly
- Step 3. Remove unneeded services from the internet and patch those that remain
- Step 4. Require MFA and watch remote access accounts
- Step 5. Limit phishing and social engineering
- Step 6. Secure Active Directory and privileged accounts
- Step 7. Limit lateral movement
- Step 8. Control which programs and scripts may run
- Step 9. Detect the attack before encryption
- Step 10. Limit the access of suppliers and managed IT providers
- How to check that the controls work
- Checklist
- A few numbers
- Where we help
- Technical basis
- Related articles
- Sources
A company can have a new threat detection system while its backups can still be deleted by anyone who takes over a domain admin account. An attacker who finds such a gap deletes the backups first and only then encrypts the data. Security deployed in a random order can leave exactly that gap open.
The #StopRansomware guide, the Mandiant M-Trends 2026 report and the Coveware report show that:
- attackers look for backups and try to delete or encrypt them, and log in to backup systems with stolen credentials,
- in 2025 exploiting a vulnerability was the most common way in, and ransomware operations most often used access taken over earlier by another group,
- in Q3 2026 ransomware attacks most often began with phishing and social engineering or with a compromise of remote access.
The steps after an attack is detected and the response plan are described in the ransomware guide.
Deployment order
The order follows three principles. First, the company must be able to restore its data without paying. Next, the entry routes that reports from 2025 and 2026 show attackers use most often must be closed. Finally, the spread of an attack must be limited and the organisation must learn to detect it.
| Step | Control | Why it sits here |
|---|---|---|
| 1 | Inventory and list of critical systems | Without an asset list nobody knows what to back up, patch and restore first. |
| 2 | Offline backups and restore tests | A backup out of the attackers’ reach decides whether the company gets back to work without a ransom. |
| 3 | Internet-facing services and their patches | According to Mandiant, exploiting a vulnerability was the most common intrusion vector in 2025. |
| 4 | MFA and control of remote access accounts | According to Coveware, remote access compromise and compromised credentials accounted for more than a third of attacks in Q3 2026. |
| 5 | Email and social engineering | According to Coveware, phishing and social engineering were the most common ransomware vector in Q3 2026. |
| 6 | Active Directory and privileged accounts | According to CISA, attackers use domain controllers to push ransomware across the whole network. |
| 7 | Limiting lateral movement | According to Coveware, lateral movement appeared in 71% of Q3 2026 cases. |
| 8 | Control over which programs and scripts can run | An allowlist blocks the tools an attacker brings into the network. |
| 9 | Detection and logs | Detection before encryption shortens the incident and limits the damage. |
| 10 | Suppliers and managed IT providers | A supplier’s access bypasses the controls of steps 3-5 unless it has its own limits. |
Steps 5-10 can run in parallel with steps 3-4 if resources allow. Steps 1 and 2 should be completed first.
Step 1. Record your assets and decide which systems are critical
The inventory covers hardware, software and data. The list of critical systems names the systems needed for the health and safety of people, the systems revenue depends on, and their dependencies. The list should record, for example, that the billing system uses a database on a specific server.
A network diagram should also be drawn and kept current: networks, addressing, connections to suppliers and to the cloud. The documentation should be stored securely, with an offline copy and a printout on site. Why: during an attack, documentation on the company network is often encrypted with everything else. For industrial networks, the inventory is covered in the article on ICS asset inventory.
Step 2. Keep offline backups and test restoring them regularly
Backups of critical data should be kept offline and encrypted. Their availability and integrity should be tested regularly in a disaster recovery scenario. Why: according to CISA, most ransomware groups look for reachable backups and delete or encrypt them.
An automatic cloud backup that synchronises files is not enough. Why: encrypted local files sync to the cloud and overwrite the good versions. Cloud backups should be protected with delete protection or object lock and versioning.
- The backup system has its own accounts, separate from the company domain. Why: CISA says attackers log in to backup systems with credentials collected in the victim’s network.
- Backup software is patched. Why: attackers target unpatched backup systems with publicly available exploits.
- The company keeps current golden images of critical systems, that is ready system templates with software for rebuilding a server quickly.
- Infrastructure-as-code (IaC) templates are versioned and copies are kept offline, so cloud resources can be redeployed.
- Offline backups include source code or installers and licence agreements, because not every image installs on new hardware.
Immutable storage protects backups from deletion, but CISA recommends using it with caution. Why: according to CISA it does not meet the criteria of some regulations, and misconfiguration can be costly.
Step 3. Remove unneeded services from the internet and patch those that remain
Remote desktop (RDP) and other remote access services should not be exposed to the internet without compensating controls. Why: according to CISA, attackers often enter through poorly secured remote services and then move through the network over RDP.
Internet-facing services should be scanned regularly with a vulnerability scanner. Patches should go first to internet-facing systems, and among those first to vulnerabilities known to be exploited. Why: according to M-Trends 2026, the mean time to exploit was about minus seven days, so exploitation began before a patch was released.
VPN gateways, network devices and the computers used for remote work should be updated first. Unneeded applications and protocols should be disabled on internet-facing systems. In industrial networks, the rules from the article on OT vulnerability management apply, because patches there need testing and maintenance windows.
Step 4. Require MFA and watch remote access accounts
MFA (multi-factor authentication) requires at least two kinds of proof of identity, such as a password and a hardware key. CISA recommends phishing-resistant MFA for all services, especially email, VPNs and accounts with access to critical systems. Systems that do not support or do not enforce MFA should be escalated to senior management.
User and administrator accounts should be reviewed quarterly, and inactive or unauthorised ones disabled. Why: at Colonial Pipeline the attackers entered through a VPN with the credentials of an old account that nobody had disabled. The review should start with remote management accounts and supplier access.
- Passwords have at least 15 characters and are unique, and the password manager has MFA enabled.
- After a few failed logins the account locks, and the logs show password guessing attempts.
- Default administrator names and passwords are changed.
- Browsers do not save passwords, enforced by group policy.
- The company monitors leaks of credentials for its domain.
Step 5. Limit phishing and social engineering
Email should be filtered so that a message with a malicious attachment does not reach the user. The mail client flags external messages, and the email gateway rejects messages with known malicious indicators. The attachment filter blocks file types that are rarely needed in email.
- The list of blocked attachment types should be reviewed at least every six months. Why: CISA gives the example of notebook files with embedded malicious code, which recently started being used in attacks.
- The company domain has SPF and DKIM records and a DMARC policy. Why: DMARC makes it harder to spoof the company domain, but protects only against senders who also use it.
- Macros in office suite files received by email are disabled, and Windows Script Host is disabled where nobody uses it.
- Training teaches staff to recognise and report phishing, including calls from people posing as IT staff.
M-Trends 2026 reports that voice phishing (vishing) was the second most common intrusion vector in 2025. CISA describes attackers who call as the help desk and trick staff into handing over credentials. The identity of anyone who calls asking for a password or MFA reset should be confirmed through another channel. Why: a phone call alone does not confirm identity. Details are in the article on email security hardening.
Step 6. Secure Active Directory and privileged accounts
According to CISA, attackers use domain controllers as the point from which they push ransomware across the whole network. Microsoft recommends the tiered administration model, which separates accounts that manage the domain from accounts that manage servers and workstations. The newer version of this model, the enterprise access model, expands tier 0 to the whole identity control plane.
Administrators have separate accounts for administration and for daily work. Domain admin accounts must not be used to log in to ordinary workstations. Why: Microsoft recommends that a lower tier must not be able to take over a higher one, and CISA describes attackers who read credentials from workstation memory.
| Control | Recommendation | Source |
|---|---|---|
| Protected Users group for high-risk accounts | High-risk accounts should be added to this group. According to CISA, the group protects against pass-the-hash attacks, that is logging in with a stolen password hash. | Microsoft, 2025; CISA |
| Service accounts as gMSA (group managed service accounts) | Microsoft recommends migrating service accounts to group managed service accounts (gMSA). | Microsoft, 2025 |
| RC4 disabled in Kerberos | Microsoft recommends disabling RC4 for Kerberos, and Windows Server 2025 disables it by default. | Microsoft, 2025 |
| KRBTGT password changed at least every 180 days, twice | According to Microsoft, only a double reset fully invalidates previously issued Kerberos tickets. | Microsoft, 2025 |
| LAPS for local administrator passwords | CISA recommends LAPS for managing local administrator passwords, and it is built into systems from Windows Server 2019 and Windows 10. | CISA |
| LSA protection and Credential Guard | These mechanisms protect against reading credentials from LSASS process memory, and Microsoft recommends LSA protection on domain controllers. | CISA; Microsoft, 2025 |
| NTLM limited or disabled | CISA recommends considering disabling or limiting NTLM, and where it must run, enabling Extended Protection for Authentication (EPA), which protects against some NTLM relay attacks. | CISA |
Domain controllers run the latest supported Windows Server version, current patches and a block on internet access. Only the necessary software and agents are installed on domain controllers. Why: according to CISA, every additional program can be used to run arbitrary code on the controller.
AD permissions should be reviewed regularly for excessive group memberships. M-Trends 2026 describes abuse of misconfigured AD certificate services templates to create administrator accounts. Details are in the articles on Active Directory misconfigurations, Kerberos attacks and NTLM relay defence.
Step 7. Limit lateral movement
According to CISA, most Windows workstations do not need access to other workstations. SMB traffic (the Windows file sharing protocol) should be limited to the systems that need it, such as workstations and domain controllers. Why: an attacker who takes over one workstation then cannot reach the others through file shares.
- SMBv1 is disabled and systems require SMB 3.1.1. Why: Microsoft rated the SMBv1 vulnerabilities in bulletin MS17-010 as critical, because they allowed code execution without authentication.
- According to CISA, requiring SMB encryption should be considered, and signing should be required when encryption is disabled.
- Local accounts cannot log in over the network, enforced by group policy.
- RDP connections to servers use Remote Credential Guard or restricted admin mode.
- Remote monitoring and management (RMM) software is inventoried, and approved tools run only from the company network over an approved VPN. RMM ports are blocked at the network perimeter.
- Hypervisors and their consoles are patched and have a hardened configuration. Why: according to CISA, taking over a hypervisor allows fast encryption of the infrastructure at scale.
The network should be divided into segments by department and function, and the OT network separated from IT. CISA warns that segmentation stops working when someone connects segments, for example with removable media. Zone design is covered in the article on OT network segmentation.
Step 8. Control which programs and scripts may run
All computers should use an application allowlist or an EDR system (endpoint detection and response) that blocks unauthorised programs. On Windows, WDAC and AppLocker do this. Why: CISA recommends an allowlist instead of trying to block every malicious program one by one.
PowerShell should be available, through group policy, only to users who need it. Older PowerShell versions should be uninstalled. Module, script block and transcription logging should be enabled, and the logs kept for at least 180 days. Why: according to CISA, attackers use PowerShell to deploy ransomware, and logs from versions older than 5.0 are not enough for analysis.
Step 9. Detect the attack before encryption
Logs from network devices, servers, workstations and cloud services should be collected centrally, for example in a SIEM, and kept for at least a year, if possible. A baseline of normal network traffic should also be set. Why: deviation from the baseline shows lateral movement and data theft before encryption starts.
| Signal | What it means | What to set up |
|---|---|---|
| New account in the domain admins group | The attacker is escalating privileges. | An alert on every change of membership in privileged groups. |
| VPN login from a new address or an old account | Someone is using compromised credentials. | An alert on logins from new locations and from accounts without activity. |
| Read of LSASS process memory | The attacker is extracting passwords from memory. | An EDR rule that blocks and reports such reads. |
| Runs of vssadmin, wbadmin, bcdedit or wmic with shadow copy commands | The attacker is deleting backups and restore points. | A high-priority alert and automatic isolation of the workstation, if the EDR allows it. |
| High outbound traffic to file storage services | The attacker is stealing data. | An alert on deviation from the outbound traffic baseline. |
| Unexpected remote management tool | The attacker is securing a way back into the network. | An alert on any RMM tool outside the allowlist. |
According to Dragos, organisations with comprehensive OT visibility detected and contained ransomware incidents in 5 days on average. The industry average was 42 days.
Step 10. Limit the access of suppliers and managed IT providers
According to CISA, attackers target managed service providers (MSPs) to reach their customers through their connections. A supplier’s security practices should be assessed before access is granted. Security requirements should be written into the contract.
- The supplier has access only to the devices and servers it is responsible for.
- If a supplier runs the backups, it follows the rules of step 2.
- Supplier access should be reviewed quarterly together with user accounts.
- In the cloud, organisation-wide policies block deleting logs and changing network configuration, even for accounts with broad permissions.
How to check that the controls work
Controls should be tested in practice, because an entry in a policy does not show whether they work. Restoring from backup should be tested on a selected critical system, measuring the time to full operation. Penetration tests and Red Team operations check the route from initial access to the domain controller and the backups.
In June 2026 NIST published the final version of the NIST IR 8374 Rev. 1 profile. The profile identifies the CSF 2.0 outcomes (the Cybersecurity Framework, NIST’s risk management framework) that support managing ransomware risk. According to NIST, the profile helps gauge readiness and prepare a countermeasure playbook.
| CSF 2.0 function | Steps in this article |
|---|---|
| Govern | Management approves the deployment order and receives reports of systems without MFA (steps 4 and 10). |
| Identify | The company knows its assets, critical systems and the vulnerabilities of internet-facing services (steps 1 and 3). |
| Protect | Backups, MFA, email, AD, segmentation and the allowlist protect the systems (steps 2, 4-8 and 10). |
| Detect | Central logs and alerts show the attack before encryption (step 9). |
| Respond and Recover | The response plan and restoration from backup are covered in the ransomware guide (step 2). |
Checklist
- The company has a current inventory, network diagram and list of critical systems, including offline versions.
- Backups of critical data are offline and encrypted, and restoring them is tested regularly.
- The backup system has its own accounts and current patches.
- No remote access service is exposed to the internet without safeguards, and internet-facing systems are patched first.
- Phishing-resistant MFA protects email, VPNs and accounts with access to critical systems.
- Accounts are reviewed quarterly, and inactive ones are disabled.
- Email has attachment filters, DMARC and macros from the internet disabled.
- Administrators have separate accounts, and AD uses Protected Users, gMSA, LAPS and KRBTGT resets.
- Domain controllers have no internet access and only the necessary software.
- SMBv1 is disabled, SMB traffic between workstations is limited, and RMM is inventoried and restricted.
- Computers use an application allowlist, and PowerShell logs are kept for at least 180 days.
- Logs are collected centrally, and alerts react to new privileged accounts, shadow copy deletion and high outbound traffic.
- Suppliers have access only to their own systems, and security requirements are in the contracts.
A few numbers
| Number | What it means | Source |
|---|---|---|
| 32% | This share of intrusions investigated by Mandiant in 2025 began with an exploited vulnerability. Email phishing accounted for 6%. | Mandiant M-Trends 2026 |
| 30% | This share of ransomware operations in 2025 began with access taken over earlier by another group, twice the 2024 share. | Mandiant M-Trends 2026 |
| about 29% | This share of ransomware attacks in Q3 2026 began with phishing and social engineering. Remote access compromise accounted for 22-23%. | Coveware, 2026 |
| 56% | This share of attacks described in the Sophos survey ended with data encrypted. The survey covered 2,158 organisations. | Sophos, 2026 |
| USD 1.7 million | This was the average cost of recovering operations after an attack, according to Sophos. | Sophos, 2026 |
Where we help
During penetration tests we check internet-facing services, the Active Directory configuration and the route to the backup systems. Red Team operations show whether the team detects an attacker before encryption. A security architect designs the deployment order and the control architecture, and in industrial plants we check the separation of IT from OT during an OT security assessment. We welcome a conversation through the contact form.
Technical basis
| Step / topic | Where it is |
|---|---|
| 1. Inventory, list of critical systems, network diagram | CISA et al., #StopRansomware Guide, part 1, General Best Practices and Hardening Guidance |
| 2. Offline backups, tests, golden images, IaC, immutable storage | CISA et al., #StopRansomware Guide, part 1, Preparing for Ransomware and Data Extortion Incidents |
| 3. Internet-facing services, scanning, patches | CISA et al., #StopRansomware Guide, Initial Access Vector: Internet-Facing Vulnerabilities and Misconfigurations; Mandiant M-Trends 2026 |
| 4. MFA, passwords, account review | CISA et al., #StopRansomware Guide, Initial Access Vector: Compromised Credentials |
| 5. Email and social engineering | CISA et al., #StopRansomware Guide, Initial Access Vector: Phishing and Advanced Forms of Social Engineering |
| 6. Active Directory | Microsoft, AD DS guidance of 9.12.2025; Microsoft, Enterprise access model; CISA et al., #StopRansomware Guide, Secure domain controllers |
| 7. Lateral movement, SMB, RMM, hypervisors, segmentation | CISA et al., #StopRansomware Guide, part 1; Microsoft, MS17-010 |
| 8. Application allowlist, PowerShell | CISA et al., #StopRansomware Guide, Initial Access Vector: Precursor Malware Infection and General Best Practices |
| 9. Logs and detection | CISA et al., #StopRansomware Guide, part 1 and the threat hunting list in part 2 |
| 10. Suppliers and MSPs | CISA et al., #StopRansomware Guide, Initial Access Vector: Third Parties and Managed Service Providers |
| Mapping to CSF 2.0 | NIST IR 8374 Rev. 1 (June 2026) |
Related articles
- Ransomware - what it is, how an attack unfolds and how to respond - the course of an attack and the steps after it is detected
- Active Directory misconfigurations - common AD gaps that attackers exploit
- OT network segmentation - dividing the network into zones and conduits
- Email security hardening - SPF, DKIM, DMARC and filters
Sources
- CISA, MS-ISAC, NSA, FBI - #StopRansomware Guide (version of 19.10.2023, PDF) - CISA (accessed: )
- M-Trends 2026: Data, Insights, and Strategies From the Frontlines - Mandiant / Google Cloud (accessed: )
- Coveware - quarterly report for Q3 2026 (28.10.2026) - Coveware (accessed: )
- Microsoft - Microsoft's guidance to help mitigate critical threats to Active Directory Domain Services in 2025 (9.12.2025) - Microsoft (accessed: )
- Microsoft - Securing privileged access: Enterprise access model - Microsoft (accessed: )
- NIST IR 8374 Rev. 1 - Ransomware Risk Management: A Cybersecurity Framework 2.0 Community Profile (June 2026) - NIST (accessed: )
- Sophos - The State of Ransomware 2026 - Sophos (accessed: )
- CyOTE - Case Study: DarkSide Ransomware Attack on Colonial Pipeline - Idaho National Laboratory / U.S. Department of Energy (accessed: )
- Microsoft Security Bulletin MS17-010 - Critical - Microsoft (accessed: )
- CrowdStrike - Big Game Hunting with Ryuk: Another Lucrative Targeted Ransomware (10.01.2019) - CrowdStrike (accessed: )
- Dragos - 2026 OT/ICS Cybersecurity Year in Review (press release of 17.02.2026) - Dragos (accessed: )
Need help in this area?
Our experts will help you assess the risk and plan next steps.