通告覆盖从边缘设备入侵到横向移动的完整链路,可直接用于暴露面自查。
勒索软件边界设备报告
Oct 21, 2026
SafePay Ransomware: SafeBreach Coverage for CISA Alert AA26-293A
Learn how SafeBreach maps CISA Advisory AA26-293A to simulations that test your exposure to SafePay ransomware’s edge-device exploitation, credential abuse, lateral movement, and data theft.
Summary
CISA Advisory AA26-293A, released October 20, 2026 by the FBI, CISA, the U.S. Secret Service (USSS), and German and UK law enforcement partners, details SafePay—active from August 2024 through 2026 and among the year’s most active ransomware groups.
SafePay has impacted over 600 commercial enterprises and critical infrastructure organizations across North America, South America, Europe, and Australia.
It is a high-complexity, multi-stage operation: edge-device CVE exploitation, credential brute-forcing, living-off-the-land binaries, and ransom.
The goal is to exfiltrate sensitive files and extort payment. The impact is mass file encryption, destruction of local and NAS backups, data theft, and public leak-site extortion.
This post covers SafePay’s four attack stages, its TTPs mapped to MITRE ATT&CK for Enterprise v20, the advisory’s IOCs, and the SafeBreach simulations—existing and newly added—that validate SafePay’s behavior from brute force through credential theft, lateral movement, exfiltration, and encryption. It also addresses two things the advisory implies but never states: why “we have MFA” does not answer this threat, and why a tested backup strategy does not close a double-extortion risk.
Author:
Tova Dvorin, Senior Product Marketing Manager
On October 20, 2026, the FBI, CISA, the U.S. Secret Service (USSS), Germany’s Landeskriminalamt Rheinland-Pfalz, the UK’s National Crime Agency, and the UK’s North West Regional Organised Crime Unit released a joint Cybersecurity Advisory titled “
#StopRansomware: SafePay Ransomware
” (Alert Code: AA26-293A).
SafePay is a double-extortion ransomware group that breaks in through VPNs, firewalls, and internet-exposed backup servers, then relies on legitimate admin tools to steal and encrypt data. The TTPs and IOCs come from FBI, USSS, and Landeskriminalamt Rheinland-Pfalz investigations as recent as August 2026—active law enforcement casework roughly two months old at publication, not indicators harvested from public reporting. Two of the six authoring bodies are investigative units rather than advisory writers.
The advisory’s Key Actions are short: patch systems, especially firewalls and edge devices; implement Zero Trust; restrict administrative tools and privileges; and strengthen account security through ICAM policies and MFA.
What is SafePay Ransomware?
SafePay is a financially motivated ransomware group first observed in August 2024. It uses double extortion—stealing data, encrypting systems, then threatening to publish on its leak site—and has impacted over 600 organizations. It typically gains initial access through VPNs, firewalls, and internet-exposed backup servers.
The group demands a subset of company revenue in cryptocurrency, leaves a ransom note on the desktop or root directory, and directs victims to a TOR .onion portal using a victim ID as the login. A representative makes phone calls to increase pressure. Non-payment means publication.
What stands out in AA26-293A is how little of this needs custom tooling. PsExec, WinRM, AnyDesk, TeamViewer, ScreenConnect, RClone, WinRAR, 7-Zip, FileZilla, NetExec, AcronisDiskDirector—built-in utilities and commercial software, across nearly every stage.
The authoring agencies call out this living-off-the-land (LOTL) approach explicitly, which means your detection surface is behavioral, not signature-based. They also caution anyone building blocklists from the IOC tables: these tools “should not be attributed as malicious without analytical evidence to support threat actor use and/or control.”
The other defining trait is where SafePay starts. Every documented initial-access route touches the network edge—a VPN, a firewall, or an internet-exposed backup server.
The 2026 Pattern: Three Advisories, One Front Door
SafePay is not doing anything novel at the perimeter. It is reusing the most reliable entry point in enterprise security, and the overlap with other recent #StopRansomware advisories is specific enough to act on.
Three of SafePay’s five CVEs also appear in the Akira advisory (
AA24-109A
, published April 2024,
updated
November 13, 2025):
Table 1: AA26-293A CVE overlap with the Akira advisory (AA24-109A)
CVE
Also exploited by
Advisory
CVE-2020-3259
Akira
AA24-109A
CVE-2023-20269
Akira
AA24-109A
CVE-2024-40766
Akira
AA24-109A (Nov 2025 update)
CVE-2023-27532
Not observed elsewhere
AA26-293A only
CVE-2024-40711
Not observed elsewhere
AA26-293A only
Akira’s advisory describes actors “obtaining initial access to organizations through a virtual private network (VPN) service without MFA configured, mostly using known Cisco product Common Vulnerabilities and Exposures (CVEs).” Its November 2025 update adds brute-forcing VPN endpoints.
Gunra (
AA26-222A
, August 10, 2026) makes three. Gunra exploited authentication bypass flaws in internet-facing firewall and VPN appliances, and Korea’s National Police Agency observed it “exploiting credential-exposure and SSH access-control weaknesses in internet-facing VPN gateways to obtain unauthorized remote access.” In one intrusion, actors reached an administrator account on an SSL-VPN appliance using default credentials, with no account lockout.
Different crews. Different encryptors. Different countries of operation. Same front door.
So a VPN and edge-device exposure review is not incident-driven work tied to one advisory—it is a standing control three separate ransomware operations have independently found profitable. If you patched for Akira in 2024 and 2025, three of SafePay’s five doors are already closed. If you did not, SafePay is the second group to walk through them.
How does a SafePay ransomware attack work?
A SafePay attack moves through four stages: initial access at the network edge, reconnaissance and privilege escalation, lateral movement and exfiltration, and destruction of recovery followed by encryption and extortion.
Stage 1: Initial Access
Attackers enter by brute-forcing exposed interfaces and running automated scripts against known edge vulnerabilities to bypass boundary authentication. The advisory adds that actors use Mullvad VPN IP addresses with phished account profiles to reach vulnerable edge devices.
Stage 2: Reconnaissance and Privilege Escalation
Actors map Active Directory network shares with ShareFinder.ps1 and SharpShares.exe, then extract administrator credentials via Mimikatz and DCSync to hold persistent control.
Stage 3: Lateral Movement and Data Exfiltration
The group moves laterally over RDP, PsExec, and WinRM, evading detection by packing executables and disabling host security software.
Stage 4: Invalidation of Recovery and Extortion
Actors dismantle recovery by terminating the Volume Shadow Copy Service and wiping NAS appliances to factory settings, then encrypt what remains, extorting victims via Tor and threatening public leak sites.
SafePay TTPs
AA26-293A maps SafePay to
MITRE ATT&CK for Enterprise v20
. The techniques below come from the advisory’s Technical Details narrative and its ATT&CK tables (Tables 6 through 16).
Initial Access
External Remote Services (
T1133
):
Actors connect over VPNs, using Mullvad VPN IP addresses and phished account profiles to exploit known edge-device CVEs:
CVE-2020-3259 (CWE-200: Exposure of Sensitive Information to an Unauthorized Actor)
CVE-2023-20269 (CWE-288: Authentication Bypass Using an Alternate Path or Channel)
Exploit Public-Facing Application (
T1190
):
Separately, actors exploit these across internet-exposed systems through automated scripts:
CVE-2023-27532 (CWE-306: Missing Authentication for Critical Function)
CVE-2024-40711 (CWE-502: Deserialization of Untrusted Data)
CVE-2024-40766 (CWE-284: Improper Access Control)
Brute Force (
T1110
):
Actors brute-force exposed interfaces and credentials. Note that while this functions as an initial-access route, AA26-293A’s ATT&CK tables map
T1110
under Credential Access (Table 12).
AA26-293A gives CWE classifications only and does not name affected vendors. Per NVD: CVE-2020-3259 and CVE-2023-20269 affect Cisco ASA and Firepower Threat Defense; CVE-2023-27532 and CVE-2024-40711 affect Veeam Backup and Replication; CVE-2024-40766 affects SonicWall SonicOS. All five are patched.
The two routes look different in your logs. The credential-led route is human-driven and resembles a legitimate remote session. The scripted route is opportunistic and resembles internet background noise until it succeeds.
Execution
PowerShell (
T1059.001
)
and
Windows Command Shell (
T1059.003
):
Actors run ShareFinder.ps1 to enumerate Active Directory file shares and identify exfiltration targets, and SharpShares.exe for the same purpose.
System Services: Service Execution (
T1569.002
):
PsExec and similar tools execute commands on remote systems.
Command and Scripting Interpreter (
T1059
):
Windows UAC elevates a program’s privileges to run malicious commands or binaries.
Persistence
Account Manipulation (
T1098
):
Actors gain administrative privileges on existing accounts.
Create Account (
T1136
):
Actors create new accounts using built-in administrative tools, consistent with LOTL. ScreenConnect also supports persistence and access.
Privilege Escalation
Bypass User Account Control (
T1548.002
):
A UAC bypass creates an elevated process through a specified component object model, running commands or binaries at administrator level.
Token Impersonation/Theft (
T1134.001
):
Actors have also used token impersonation.
Stealth and Defense Impairment
Disable or Modify Tools (
T1685
):
Before moving laterally, actors disable security software, modify or delete registry keys and configuration files, and disable security patches.
Software Packing (
T1027.002
):
Packing or virtual machine protection conceals their code.
System Binary Proxy Execution (
T1218
):
Execution is proxied through trusted, often Microsoft-signed binaries to bypass signature-based defenses. The advisory names PsExec and WinRM.
File Deletion (
T1070.004
):
After encryption, SafePay may delete itself.
Credential Access
Brute Force (
T1110
):
Brute-force password attacks against accounts.
Mimikatz harvests authentication credentials including Kerberos tickets
(
T1550.002
)
. The IOC list includes dcsync.exe, “a tool used to perform DCSync operations to extract Active Directory credentials; high-confidence credential theft activity.”
In some cases actors changed system user passwords, making valid logins impossible.
Discovery
Actors use SharpShares.exe, ShareFinder.ps1, netscan.exe, NetExec (nxc.exe), and nltest.exe to find hosts, shares, and domain trusts. The IOC list also includes their output artifacts—domains.txt, oui.txt, ping.txt, scan.xml, subs.txt—which are hunting targets in their own right.
Lateral Movement
Remote Desktop Protocol (
T1021.001
):
Actors use RDP with valid accounts and typically disable Windows Defender while using LOLBins.
SMB/Windows Admin Shares (
T1021.002
):
SMB carries lateral movement through the network.
Windows Remote Management (
T1021.006
):
WinRM executes commands on remote systems.
Remote Services (
T1021
):
Remote services generally expand their presence.
Collection
Archive via Utility (
T1560.001
):
FileZilla, 7-Zip, RClone, and WinRAR compress and collect data. WinRAR splits data into segments and compresses to .RAR before exfiltration.
Archive via Library (
T1560.002
):
Mapped by the advisory to credential dumpers such as Mimikatz.
Exfiltration
Over Alternative Protocol (
T1048
):
WinSCP and RClone move data to actor-controlled accounts; RClone syncs to cloud storage such as Mega. Actors have also run b2.exe to exfiltrate to Backblaze B2.
Over Web Service (
T1567
):
OneDrive, in some instances.
Actors install AnyDesk or TeamViewer on victim systems to transfer ransomware and other software in.
Impact
Inhibit System Recovery (
T1490
):
Actors halt the Volume Shadow Copy Service and reset NAS holding backup data to factory settings. The ATT&CK table goes further: in some cases they “deleted all network-attached backups of company data.”
Data Encrypted for Impact (
T1486
):
A hybrid ChaCha20 and RSA scheme encrypts files, varying method by file type and size and supporting full or partial encryption. Files are appended with .SafePay and cannot be recovered without the key.
What Third-Party Tools Does SafePay Abuse?
Three tools carry most of the weight:
Mimikatz:
Harvested plaintext credentials, password hashes, and Kerberos tickets from memory to facilitate privilege escalation.
PsExec:
Executed remote commands and launched secondary payloads across compromised endpoints via Windows service execution.
SharpShares:
Conducted automated network host reconnaissance to discover accessible shares and map enterprise storage.
The advisory’s Table 1 also lists 7-Zip, FileZilla, PowerShell, RClone, ScreenConnect, ShareFinder.ps1, WinRAR, WinRM, AnyDesk and TeamViewer, AcronisDiskDirector, NetExec, and Backblaze (b2.exe).
SafePay IOCs
IOCs observed between November 2024 and August 2026:
Nine IP addresses
—victim intrusions (104.26.2[.]223, 140.82.114[.]4, 185.199.108[.]133, 77.37.49[.]40), hosting victim data (173.194.194[.]94), client login panels (167.235.225[.]149), exfiltration (5.255.11[.]134, 37.10.71[.]188), and soc.dll command and control (88.119.167[.]239)
Roughly 30 filenames
, including locker.dll, dcsync.exe, Socks.exe, Socks_werfault.exe, sharefinder.ps1, netscan.exe, nxc.exe, b2.exe, DYRasWtb.exe, and netscanNEW.zip, plus legitimate binaries that are suspicious in context (CCleaner.exe, svchost.exe outside System32, nltest.exe outside admin workflows)
Four SHA-1 and six SHA-256 hashes
for the ransomware binary, locker.dll, soc.dll, SharpShares.exe, NetExec, a RunPE bundled DLL, and an injected PE
Four actor hostnames:
WIN-SBOE3CPNAL, WIN-3IUUOFVTQAR, WIN-DK18BI83M0S, WIN-I1HRRJHCEIV
Download the AA26-293A
STIX XML
or
STIX JSON
from CISA for the full list.
Two cautions govern how you use them. Verify IP addresses before blocking. And because malicious infrastructure often sits on cloud platforms with dynamically assigned addresses, “an address associated with malicious activity at one point in time may later be reassigned to benign services”—treat these as historically observed infrastructure, corroborated against current telemetry.
This is the argument for detecting behavior rather than indicators alone. Addresses and hashes age in days. A crew’s preference for PsExec, RClone, ShareFinder, and WinRAR has held across two years of casework.
“We Have MFA” Is Not an Answer to This Threat
The advisory’s MFA guidance reads like a box already ticked. Most organizations running a VPN have MFA on it. SafePay still got into more than 600 of them.
The gap shows in the mitigation language. The agencies recommend ICAM policies and MFA “for all services to the greatest extent possible,” then separately require all accounts with password logins—naming service accounts, admin accounts, and domain admin accounts—to meet NIST’s digital identity guidelines. They also recommend reviewing domain controllers, servers, workstations, and Active Directory “for new, unactive and/or unrecognized accounts.”
Three recommendations, because enforcement coverage and policy coverage are different things. The accounts that break the pattern are familiar to anyone who has run an identity program:
Break-glass accounts, deliberately exempted and rarely reviewed
Service accounts authenticating to the same gateway as humans
Local accounts on the appliance itself, outside the identity provider that enforces your policy
Contractor accounts from a project that ended, still enabled
Accounts an attacker already inside created or elevated—SafePay’s
T1098
and
T1136
activity means last quarter’s inventory is not today’s
SafePay’s initial access compounds this. Phished account profiles over Mullvad VPN IPs defeat SMS and push-based MFA, which is why Akira’s guidance specifies phishing-resistant factors. And brute force only works where lockout thresholds are absent, which is why the advisory separately recommends failed-login lockouts.
The question that produces an answer is not “do we require MFA on the VPN.” It is: enumerate every identity that can authenticate to an internet-facing service, and name the factor for each. The ones with no factor are your exposure. That list is usually short, surprising, and fixable in an afternoon.
Backups Answer Encryption. They Don’t Answer Extortion.
The reflex response to a ransomware advisory is to confirm the backup strategy. AA26-293A gives two reasons that is not enough.
First, the backups are a target.
SafePay halts the Volume Shadow Copy Service and wipes NAS holding backup data to factory settings; the ATT&CK table states that in some cases actors deleted all network-attached backups. Two of the five CVEs—CVE-2023-27532 and CVE-2024-40711—hit backup infrastructure directly. A backup server reachable from the internet is not a recovery control; it is an initial access vector holding your recovery data. Hence the advisory’s insistence on offline, encrypted, immutable backups in a physically separate, segmented location—and why “we back up nightly” and “we can recover” are different claims.
Second, recovery does not undo disclosure.
SafePay steals data before encrypting it. Restoring from a clean backup returns your systems. It does not un-publish the leak site, stop the pressure calls the advisory describes, or remove your notification obligations to regulators, customers, and employees. A company that restores perfectly and pays nothing has still suffered a data breach.
The practical test: does your incident response plan have a path for “data stolen, systems still running”? Most have a restore runbook and an outage comms tree. Far fewer have a decision tree for the window between exfiltration and encryption, where detection is still possible and the outcome still changeable. That window is where the advisory’s exfiltration controls live: segmentation, filtering untrusted origins from remote services, and network monitoring that logs lateral movement.
For boards, the reframe is simple. Backups are an availability control. Double extortion is a confidentiality event. Different investments answer them, and only one is usually funded.
SafeBreach Coverage & Playbook Attack Updates
Existing Behavioral Coverage
SafeBreach provides simulations for much of this campaign. These are mapped to the ATT&CK techniques AA26-293A names and can be run now. Several, including #6513, #8857, #11773, and #7165, are Advanced Actions and should be scheduled accordingly.
T1110
/
T1021.001
—Brute force and RDP
#192—Brute force attack over RDP protocol
#11773—Brute force attack over RDP protocol against a live service (Windows)
#6909—RDP Connection Between 2 Simulators
#6473—Agentless lateral movement via RDP
T1021.002
/
T1569.002
/
T1134.001
/
T1550.002
—PsExec and SMB
#1339—PsExec remote command execution traffic simulation
#6513—Agentless lateral movement via SMB and RCE, using Mimikatz
#6483—Agentless lateral movement via SMB and RCE, using Token duplication
#6550—Agentless lateral movement via SMB and RCE, using Impersonated user
#2273—Pass the Hash over SMB using Mimikatz
T1021.006
—WinRM
#5671—Agentless lateral movement via Windows Remote Management
#5674—Agentless lateral movement via Remote PowerShell (WinRM)
T1059.001
/
T1059.003
—Share hunting and Mimikatz
#11542—Network share discovery using net.exe (Windows)
#1220—Inject Mimikatz using PowerShell to Extract Credentials
T1098
/
T1136
—Accounts
#2189—Account Manipulation
#2170—Create Local Account (Windows)
#7170—Add a local administrator (Windows)
#10794—Create Domain Account (Windows)
T1685
—Disable or Modify Tools (Defender)
#7834—Add Exclusions to Windows Defender
#7835—Disable Windows Defender From Registry
#8855—Add exclusion to Windows Defender via writing to registry
#8854—Revert Windows Defender signatures
#8857—Add Mimikatz as exclusion to Windows Defender and execute Mimikatz
T1027
/
T1070.004
/
T1560.001
—
Obfuscation, indicator removal, and staging
#3829—Run obfuscated Mimikatz on host
#2245—Indicator Removal on Host (Windows)
#806—Collect Files using ZIP
T1048
/
T1567
—Exfiltration
#103—Data asset exfiltration using FTP STOR (ASCII)
#104—Data asset exfiltration using FTP STOR (Binary)
#122—Data asset exfiltration over SSH
#6975—Covert archived data asset exfiltration using HTTP/S POST
#7200—Covert data asset exfiltration using HTTP/S POST file upload
#8899—Sync a file outside of the OneDrive folder
T1490
/
T1486
:
Recovery tampering and encryption
#6372—Modify Volume Shadow Copy (VSS)
#7165—Execute generic ransomware from a trusted path on user files
Previously Covered IOCs
One of the advisory’s hash-identified tools was already in the playbook before AA26-293A published. SharpShares.exe—the share enumeration tool from Stage 2—matches the advisory’s SHA-256 ending bc6f41:
#10707—Write SharpShares (bc6f41) hacktool to disk
#10708—Pre-execution phase of SharpShares (bc6f41) hacktool (Windows)
#10709—Transfer of SharpShares (bc6f41) hacktool over HTTP/S
#10710—Transfer of SharpShares (bc6f41) hacktool over HTTP/S
#10711—Email SharpShares (bc6f41) hacktool as a compressed attachment
#10712—Email SharpShares (bc6f41) hacktool as a compressed attachment
Customers who have run these already have results against a named indicator.
New IOC-Based Coverage
We have added coverage for the SafePay locker payload, using the advisory’s locker.dll PE32 executable (SHA-256 ending f67917):
#11836—Write locker_dll_ (f67917) ransomware to disk
#11837—Transfer of locker_dll_ (f67917) ransomware over HTTP/S
#11838—Transfer of locker_dll_ (f67917) ransomware over HTTP/S
#11839—Email locker_dll_ (f67917) ransomware as a compressed attachment
#11840—Email locker_dll_ (f67917) ransomware as a compressed attachment
Together, the behavioral and IOC-based simulations cover SafePay’s reconnaissance, credential access, privilege escalation, defense evasion, lateral movement, staging, exfiltration, payload delivery, and impact.
Coverage Scope
These simulations validate SafePay’s brute force, credential theft, lateral movement, defense tampering, staging, exfiltration, and impact behaviors, including Volume Shadow Copy modification and real encryption of user files on the target simulator, along with the file-based indicators for SharpShares and the SafePay locker.
How to Test Your Defenses Against
SafePay Ransomware with SafeBreach
SafeBreach customers can now validate their security controls against these TTPs in multiple ways.
Method 1
– SafeBreach Scenarios:
Navigate to the SafeBreach Scenarios page and choose CISA Alert AA26-293A.
Method 2
–
Attack Playbook:
From the Attack Playbook, filter for the attack IDs listed in SafeBreach Coverage above.
Method 3 — Known Threats Series:
Open the Known Threats Series report, select the AA26-293A report, and run the simulations.
Additional Advisory Steps
Run the SafeBreach Platform Simulations
Open the updated playbooks mapped to AA26-293A.
Run simulations covering brute force, credential theft, lateral movement with PsExec and WinRM, Defender tampering, staging and exfiltration, and the locker payload IOCs.
Start with #11836, #11773, #6513, #8857, #7200, and #7165 for coverage from payload delivery through encryption.
Review results to validate detection, surface gaps, and guide remediation.
Mitigation Strategies
The authoring agencies recommend:
Remediate known exploited vulnerabilities (KEVs)
and keep systems, software, and firmware current, prioritizing internet-facing systems and the five CVEs in AA26-293A. CISA’s Binding Operational Directive 26-04 frames patch prioritization by technology type and forensics triage.
Implement ICAM policies and require MFA
across all services to the greatest extent possible, particularly webmail, VPNs, and accounts reaching critical systems.
Hold all password-login accounts
—service, admin, and domain admin included—to NIST’s digital identity guidelines: 15 to 64 characters, hashed in a recognized password manager, salted on shared credentials, never reused, with failed-login lockouts, no password hints, and no forced rotation more than annually.
Maintain offline backups
and test restoration regularly.
Keep backup data encrypted and immutable
across the entire data infrastructure.
Run a recovery plan
retaining multiple copies of sensitive data in a physically separate, segmented, secure location.
Segment networks
to control traffic between subnetworks and restrict lateral movement.
Monitor for abnormal activity
with a tool that logs and reports all network traffic, including lateral movement. EDR is particularly useful for lateral connections.
Filter network traffic
, blocking unknown or untrusted origins from internal remote services.
Review domain controllers, servers, workstations, and Active Directory
for new, unactive, or unrecognized accounts.
Disable unused ports
and consider an external-email banner.
Audit administrative accounts
and apply least privilege.
Use time-based access
for admin-level accounts and higher, such as Just-in-Time, supporting least privilege and the Zero Trust Maturity Model.
Disable command-line and scripting permissions
, since escalation and lateral movement depend on them.
If You Detect a Compromise
If compromise is detected before encryption, the agencies recommend:
Isolate compromised hosts by quarantining or taking them offline.
Threat hunt to scope the intrusion: artifacts, logs, TTPs, compromised devices and accounts, timeline.
Report to the FBI and other agencies as appropriate. In the U.K., report to Report Fraud or a local police force.
Apply eviction countermeasures with CISA’s Eviction Strategies Tool—Playbook-NG and the COUN7ER countermeasure database—once hunting data can inform the selection.
Harden the network against further activity.
If encryption has already happened, see the “Ransomware and Data Extortion Response Checklist” in CISA’s #StopRansomware Guide. The agencies do not encourage paying: payment guarantees nothing and may embolden adversaries to target others.
Proactive Threat Monitoring
Remote access tools:
Flag unapproved AnyDesk, TeamViewer, and ScreenConnect installations, especially on servers.
Staging and transfer tools:
Alert on RClone, WinSCP, FileZilla, 7-Zip, WinRAR, and b2.exe, especially uploads to Mega, Backblaze B2, or OneDrive.
Account changes:
Investigate new accounts, privilege grants, and unexpected password resets on domain controllers. The advisory notes actors changed passwords to lock users out.
Recovery tampering:
Alert on Volume Shadow Copy shutdowns and changes to network-attached backup storage. Treat a NAS factory reset as a security incident, not a hardware event.
VPN logins from anonymizing services:
Review authentications from Mullvad and other commercial VPN ranges.
Outbound transfers:
Identify which hosts may open outbound SFTP, FTP, or HTTPS to arbitrary destinations. A file server making a transfer it has never made before is the clearest signal of exfiltration in progress.
Stay Ahead with SafeBreach
For a full view of your gaps against SafePay, sign into the SafeBreach platform and run the simulations mapped to AA26-293A. You can go further with
SafeBreach Propagate
, which enables you to assess how attackers could pivot across your environment post-compromise. It maps high-risk attack paths, visualizes lateral movement, and prioritizes remediation efforts based on exposure to your most critical assets. Learn more in the
SafeBreach Propagate solution brief
or
schedule a demo
.
About the Author
Tova Dvorin
is currently the Senior Product Marketing Manager at SafeBreach. She has 8 years of Product Marketing experience in the cybersecurity sector, after a previous career as a war journalist.
Frequently Asked Questions
What is CISA Advisory AA26-293A?
A joint #StopRansomware advisory released October 20, 2026 by the FBI, CISA, the U.S. Secret Service, Germany’s Landeskriminalamt Rheinland-Pfalz, the UK’s National Crime Agency, and the UK’s North West Regional Organised Crime Unit. It details SafePay ransomware TTPs and IOCs from investigations as recent as August 2026. SafePay has impacted over 600 commercial enterprises and critical infrastructure organizations since August 2024.
Which vulnerabilities does SafePay ransomware exploit?
Five, all patched. CVE-2020-3259 and CVE-2023-20269 are exploited with phished account profiles over Mullvad VPN IP addresses; CVE-2023-27532, CVE-2024-40711, and CVE-2024-40766 are exploited across internet-exposed systems through automated scripts. The advisory lists CWE classifications rather than product names; per NVD, these affect Cisco ASA and Firepower Threat Defense, Veeam Backup and Replication, and SonicWall SonicOS. SafePay also brute-forces exposed interfaces.
How does SafePay ransomware relate to Akira and Gunra?
Three of SafePay’s five CVEs—CVE-2020-3259, CVE-2023-20269, and CVE-2024-40766—also appear in the Akira advisory,
AA24-109A
. The Gunra advisory,
AA26-222A
, describes authentication bypass flaws in internet-facing firewall and VPN appliances and credential-exposure weaknesses in VPN gateways. All three exploit the same class of entry point: internet-facing remote access with incomplete MFA enforcement and unpatched edge devices.
Do backups protect against SafePay ransomware?
They address encryption, not extortion. SafePay steals data before encrypting and publishes it if the victim does not pay, so restoring does not remove the disclosure or the notification obligations. SafePay also targets backups directly—halting the Volume Shadow Copy Service, wiping NAS to factory settings, and exploiting two CVEs in backup software. AA26-293A recommends offline, encrypted, immutable backups in a physically separate, segmented location.
Which legitimate tools does SafePay ransomware abuse?
Most central are Mimikatz for credential harvesting, PsExec for remote execution, and SharpShares for reconnaissance. AA26-293A also lists WinRM, AnyDesk, TeamViewer, ScreenConnect, 7-Zip, WinRAR, ShareFinder.ps1, netscan.exe, NetExec, nltest.exe, RClone, WinSCP, FileZilla, OneDrive, and Backblaze B2. The advisory cautions against attributing these as malicious without analytical evidence, so restrict and monitor them rather than blocking outright.
How can organizations test their exposure to SafePay ransomware?
The advisory’s Validate Security Controls section recommends testing your security program against its ATT&CK-mapped behaviors, continually and at scale in production. The SafeBreach CTEM Platform allows customers to select CISA Alert AA26-293A on the Scenarios page, filter the Attack Playbook by AA26-293A, or open the Known Threats Series report and run the simulations. Results show whether existing controls detect and block SafePay’s brute force, credential theft, lateral movement, exfiltration, and encryption behaviors.
You Might Also Be Interested In
Research
Blog
Gunra Ransomware: SafeBreach Coverage for CISA Alert AA26-222A
Read More
Solution Brief
SafeBreach CTEM Platform
Read More
Research
Blog
SafeBreach Coverage for US CERT AA25-050A [Ghost (Cringe) Ransomware]
Read More
Get the latest
research and news