Threat Overview
Google Threat Intelligence Group and Mandiant published new research on BREEZE COMET, formerly UNC5669, a financially motivated threat actor targeting Brazilian financial services, fintech, retail, e-commerce, government, and banking environments. The actor has used password spraying, IT-support vishing, RMM software, compromised trusted government websites, service-account abuse, RDP, SMB, custom tunneling, Kubernetes persistence, custom Java/Nim/Golang malware, firewall changes, scheduled tasks, and event-log clearing. The group ultimately targets payment systems and APIs to execute fraudulent transactions and has shown signs of expanding infrastructure beyond Brazil.
References
- Google Threat Intelligence Group / Mandiant, September 1, 2026: https://cloud.google.com/blog/topics/threat-intelligence/financially-motivated-threat-actor-breeze-comet-targets-brazil
- Dark Reading, September 3, 2026: https://www.darkreading.com/threat-intelligence/breeze-comet-brazilian-global-financial-systems
Impacted Systems
Targeted environments: financial services, fintech, retail, e-commerce, banks, payment processors, government, and organizations with access to banking/payment APIs. Platforms observed include Windows, Active Directory, cloud environments, Kubernetes, and enterprise networks. Initial access has included password spraying, vishing, RMM installation, compromised trusted websites, and rogue physical network devices. Post-compromise tooling includes AnyDesk, XWORM, COBALTSPIN, LIGHTPAINT, MILDFROST, KICKPLATE, and BOATBEAM.
Why this matters
The campaign is an active financially motivated intrusion set with a clear path from identity compromise to remote access, lateral movement, persistence, tunneling, and fraudulent payment activity. Its behaviors generalize well to MSSP detection across customers even outside Brazil.
Criticality: High
Exploitation Status
Confirmed active threat activity based on multiple Mandiant investigations beginning in 2024 and continuing through 2026. GTIG reports successful fraudulent-transfer activity and ongoing development.
What this hunt looks for
Available Microsoft Sentinel telemetry can be reviewed for password spraying, RMM execution, RDP and SMB activity, scheduled task/startup persistence, firewall rule modification, event-log clearing, and suspicious tunneling or rare outbound connectivity where the relevant sources are ingested.
Required logs
- Microsoft Defender XDR / Defender for Endpoint tables: DeviceProcessEvents and DeviceNetworkEvents
- Microsoft Entra ID sign-in telemetry in SigninLogs
- Windows Security Events in SecurityEvent, including 4624 and 1102
First Pass – RMM and Vishing Follow-On Activity
DeviceProcessEvents
| where TimeGenerated > ago(30d)
| where FileName in~ ("AnyDesk.exe", "ScreenConnect.ClientService.exe", "TeamViewer.exe", "rustdesk.exe", "splashtop-streamer.exe") or ProcessCommandLine has_any ("AnyDesk", "ScreenConnect", "TeamViewer", "RustDesk", "Splashtop")
| project TimeGenerated, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName, SHA1
| order by TimeGenerated desc
Password Spray and Multi-IP Authentication Failures
SigninLogs
| where TimeGenerated > ago(14d)
| where ResultType != 0
| summarize Attempts=count(), Users=dcount(UserPrincipalName), IPs=dcount(IPAddress), UserList=make_set(UserPrincipalName,20) by IPAddress, bin(TimeGenerated, 15m)
| where Attempts >= 20 and Users >= 5
| order by TimeGenerated desc
Suspicious RDP From New Sources
SecurityEvent
| where TimeGenerated > ago(14d)
| where EventID == 4624 and LogonType == 10
| summarize FirstSeen=min(TimeGenerated), LastSeen=max(TimeGenerated), Hosts=dcount(Computer), Events=count() by Account, IpAddress
| order by LastSeen desc
SMB and Network Discovery Tooling
DeviceProcessEvents
| where TimeGenerated > ago(14d)
| where ProcessCommandLine has_any ("net view", "net use", "nltest", "dsquery", "arp -a", "route print", "Get-SmbShare", "Get-NetTCPConnection") or FileName in~ ("net.exe", "nltest.exe", "dsquery.exe")
| project TimeGenerated, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName
| order by TimeGenerated desc
Scheduled Task and Startup Persistence
DeviceProcessEvents
| where TimeGenerated > ago(14d)
| where FileName in~ ("schtasks.exe", "reg.exe", "powershell.exe", "cmd.exe")
| where ProcessCommandLine has_any ("/create", @"\Run", @"\RunOnce", "Startup", "schtasks")
| project TimeGenerated, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName
| order by TimeGenerated desc
Defender Firewall Rule Modification
DeviceProcessEvents
| where TimeGenerated > ago(14d)
| where FileName in~ ("netsh.exe", "powershell.exe", "pwsh.exe")
| where ProcessCommandLine has_any ("advfirewall", "New-NetFirewallRule", "Set-NetFirewallRule", "firewall add rule")
| project TimeGenerated, DeviceName, AccountName, FileName, ProcessCommandLine, InitiatingProcessFileName
| order by TimeGenerated desc
Event Log Clearing
SecurityEvent
| where TimeGenerated > ago(14d)
| where EventID == 1102
| project TimeGenerated, Computer, Account, SubjectUserName, SubjectDomainName
| order by TimeGenerated desc
Suspicious Outbound Tunneling From Compromised Hosts
DeviceNetworkEvents
| where TimeGenerated > ago(14d)
| where InitiatingProcessFileName in~ ("java.exe", "java", "powershell.exe", "pwsh.exe", "ssh.exe", "plink.exe", "socat", "cloudflared.exe")
| summarize FirstSeen=min(TimeGenerated), LastSeen=max(TimeGenerated), Connections=count(), Ports=make_set(RemotePort,20) by DeviceName, InitiatingProcessFileName, RemoteUrl, RemoteIP
| order by Connections desc
Detection Notes
Highest signal comes from unexpected RMM usage, new RDP sources, SYSTEM scheduled tasks, firewall rule changes, event-log clearing, and tunneling processes on endpoints that do not normally perform those actions. Exact custom-malware filenames and IOCs can change, so behavior should remain primary. Financial transaction fraud itself may not be visible in Sentinel unless payment/application telemetry is ingested.