Cyber Threat Brief — February 24, 2026

⚠️ This report is AI-generated. Always validate findings.

Cyber Threat Brief — February 24, 2026

Window: February 23–24, 2026 | Audience: Detection Engineers
TL;DR: CrowdStrike drops its annual report (29-min breakout, AI at scale), Huntress dissects a ClickFix → Matanbuchus 3.0 → AstarionRAT pre-ransomware chain with 40-min DC compromise, Silver Fox abuses fake Huorong AV to drop ValleyRAT via DLL sideloading, APT28 weaponizes webhook URLs as Office macro C2, seven MCP servers get RCE CVEs from the same eval()/exec() root cause, Dragos names three new OT threat groups plus 3,300 industrial ransomware victims, and DFIR Report details an ActiveMQ exploit leading to LockBit deployment with a 90-minute final detection window.


1. CrowdStrike 2026 Global Threat Report: AI Is Both Weapon and Target

What’s New

CrowdStrike released its 2026 Global Threat Report today, tracking 280+ named adversaries in 2025. The headline numbers: eCrime average breakout time crashed to 29 minutes (fastest observed: 27 seconds), AI-enabled adversary activity surged 89% year-over-year, and 42% of vulnerabilities were weaponized before public disclosure.

Technical Details

FieldValue
Adversaries tracked280+ named
Avg. eCrime breakout29 min (65% faster than 2024)
Fastest breakout27 seconds
Earliest exfil observed4 min post-access
AI-enabled activity increase+89% YoY
Zero-days exploited pre-disclosure42% of total vulns
Cloud intrusions+37% (state-nexus +266%)
China-nexus activity+38%; logistics +85%
DPRK incidents+130% (FAMOUS CHOLLIMA doubled)
Largest single heist$1.46B crypto theft (PRESSURE CHOLLIMA)

TTPs (Key Themes)

TacticTechniqueObservable
Initial AccessT1566 – PhishingPrompt injection into GenAI tools at 90+ orgs
Credential AccessT1110 – Brute ForceAI-generated scripts for credential dumping
Defense EvasionT1070 – Indicator RemovalAI erasing forensic evidence (PUNK SPIDER)
Command & ControlT1071 – App Layer ProtoMalicious prompts via GenAI as C2 channel
ImpactT1486 – Data EncryptedPre-positioned cloud access → ransomware

Detection Opportunities

  • Prompt Injection: Monitor GenAI tool audit logs for anomalous system prompts requesting credential output or command execution
  • AI Dev Platform Abuse: Baseline API usage from AI development tooling; alert on persistence mechanisms (scheduled tasks, cron) created from AI agent processes
  • Sub-27-Second Breakout: Tune lateral movement detection for initial-to-pivot under 60 seconds; traditional 5-minute correlation windows miss this
  • Cloud Targeting: Alert on new cross-cloud IAM role assumptions, especially by identities that haven’t previously accessed target accounts

Log Sources

  • GenAI platform audit logs (OpenAI, Gemini, Copilot Enterprise)
  • Cloud provider IAM event logs (CloudTrail, Azure Monitor, GCP Audit)
  • EDR telemetry for sub-2-minute process chains
  • Network: anomalous TLS connections from AI service processes

Detection Coverage

SourceStatus
Sigma❌ Gap — no GenAI abuse or 27-sec breakout rules
Splunk ESCU❌ Gap
Elastic✅ Partial — defense_evasion_genai_process_encoding_prior_to_network_activity.toml + 3 related GenAI rules

Recommendation: Review Elastic’s GenAI detection rules as a starting baseline. AI platform audit logging is the critical gap — most orgs have it unconfigured.

Sources


2. ClickFix + Matanbuchus 3.0 + AstarionRAT: Pre-Ransomware Loader Chain

What’s New

Huntress documented a February 2026 intrusion where ClickFix delivered Matanbuchus 3.0 (a $10K–$15K/month MaaS loader), which staged a never-before-documented RAT (AstarionRAT) before the operator moved laterally to both domain controllers in under 40 minutes. This is the most technically detailed ClickFix-to-ransomware-precursor chain publicly documented to date, revealing 6+ layered stages.

Technical Details

FieldValue
Initial vectorClickFix social engineering
First stageSilent MSI install (-PaCkAGe hxxp://binclloudapp[.]com/466943 /q)
Case mixing evasionmSiexeC.EXe — mixed case to evade string-match rules
LoaderMatanbuchus 3.0 (complete rewrite; $10K/mo HTTPS, $15K/mo DNS)
Second RATAstarionRAT — 24 commands, RSA C2, SOCKS5, reflective loading
DLL Sideloading viaZillya! Antivirus AVCore.exeSystemStatus.dll (Matanbuchus)
Staging dirC:\ProgramData\USOShared\ (mimics Windows Update)
Lateral movementPsExec to 2 DCs within 40 min; rogue account creation
Defender evasionDefender exclusions staged pre-movement
C2 masqueradeRSA-encrypted traffic disguised as application telemetry
Key IOCsbinclloudapp[.]com (reg. Feb 5), IP 192.121.23[.]146 (M247 EU AS9009)

TTPs

TacticTechniqueObservable
Initial AccessT1204.002 – User Execution: Malicious FileClickFix clipboard paste → msiexec
Defense EvasionT1036 – MasqueradingMixed-case mSiexeC.EXe in command line
Defense EvasionT1574.002 – DLL Side-LoadingAVCore.exe loads malicious SystemStatus.dll
Defense EvasionT1562.001 – Disable/Modify ToolsDefender exclusion added for process + dir
PersistenceT1053.005 – Scheduled TaskPre-staged in C:\ProgramData\USOShared\
Credential AccessT1003 – OS Credential DumpingAstarionRAT credential theft module
Lateral MovementT1021.001 – RDP / T1569.002 – PsExecLateral to domain controllers
C2T1095 – Non-Application Layer ProtocolCustom Lua 5.4.7 interpreter + reflective PE loader

Detection Opportunities

# Detect mixed-case msiexec (case-insensitive matching)
CommandLine|re: '(?i)ms[Ii][Ee][Xx][Ee][Cc].*(/|\\-)[Pp][Aa][Cc][Kk][Aa][Gg][Ee]'

# Detect silent MSI install from new/untrusted domain
process.name = "msiexec.exe" AND 
CommandLine CONTAINS "/q" AND 
CommandLine MATCHES "hxxp://" AND
NOT (domain IN known_software_update_domains)

# Defender exclusion staging before lateral movement
process.name IN ("powershell.exe", "cmd.exe") AND
CommandLine CONTAINS ("Add-MpPreference" OR "Set-MpPreference") AND
CommandLine CONTAINS ("-ExclusionPath" OR "-ExclusionProcess") AND
NOT parent.name = "svchost.exe"

# AstarionRAT C2 - RSA-encrypted traffic disguised as telemetry
# Hunt for non-browser processes with high-entropy TLS to newly registered domains

# PsExec to multiple targets within 40 minutes
parent.name = "psexec.exe" AND 
dest.is_domain_controller = true AND
count(unique dest) >= 2 WITHIN 40min

# USOShared staging directory abuse
file.path CONTAINS "C:\\ProgramData\\USOShared\\" AND
file.extension IN (".dll", ".exe") AND
NOT process.name IN ("usocoreworker.exe", "musnotification.exe")

Log Sources

  • Sysmon Events 1 (process creation), 7 (image load), 11 (file create), 12/13 (registry)
  • Windows Security Event 4688 (process creation with command line)
  • EDR telemetry (command-line visibility critical)
  • Network: TLS fingerprinting for non-standard JA3 from msiexec.exe

Detection Coverage

SourceStatus
Sigma✅ ClickFix: proc_creation_win_susp_clickfix_filefix_execution.yml
Sigma✅ MSI network: net_connection_win_msiexec_http.yml
Sigma✅ DLL sideloading: image_load_side_load_from_non_system_location.yml
Sigma✅ PsExec: 36 existing rules
Sigma❌ Gap: Matanbuchus 3.0 loader specifics
Sigma❌ Gap: AstarionRAT C2 pattern (RSA telemetry masquerade)
Sigma❌ Gap: USOShared staging directory abuse
Splunk ESCU❌ Gap: Matanbuchus/AstarionRAT
Elastic❌ Gap

Write this rule: Sigma for USOShared directory PE drops from non-Windows-Update processes. Also: msiexec mixed-case detection.

Sources


3. Silver Fox APT: Fake Huorong AV Site Delivers ValleyRAT/Winos4.0 via DLL Sideloading

What’s New

Silver Fox APT (Chinese-speaking) cloned the popular Huorong Security (火绒) antivirus website — registered huoronga[.]com — to deliver ValleyRAT, a modular RAT built on the Winos4.0 framework. Malwarebytes disclosed the full technical chain today. Key detection angle: WavesSvc64.exe with no audio stack loaded, scheduled task named “Batteries,” and Winos4.0 C2 at 161.248.87.250:443 using a custom non-TLS binary protocol.

Technical Details

FieldValue
AttributionSilver Fox APT (Chinese-speaking)
DeliveryTyposquatted site huoronga[.]com
Relay domainhndqiuebgibuiwqdhr[.]cyou
Payload hostCloudflare R2 (trusted CDN)
ArchiveBR火绒445.zip
Installer typeNSIS (Nullsoft) — legitimate framework
Loader binaryWavesSvc64.exe (audio service masquerade)
DLL sideloadedDuiLib_u.dll (hijacked DirectUI library)
Encrypted shellcodebox.ini
PersistenceScheduled task “Batteries” (C:\Windows\Tasks\Batteries.job)
Registry C2 configHKCU\SOFTWARE\IpDates_info (encoded domain)
C2yandibaiji0203[.]com + 161.248.87.250:443 (custom binary proto)
Defender bypassPowerShell Add-MpPreference high-integrity exclusion for %APPDATA%\trvePath + WavesSvc64.exe
File refreshPeriodic delete-rewrite of DLLs to defeat hash-based detection

TTPs

TacticTechniqueObservable
Initial AccessT1192 – Spearphishing Link / T1036 – MasqueradeTyposquatted AV download site
ExecutionT1204.002 – User ExecutionNSIS installer mimics legitimate AV setup
Defense EvasionT1574.002 – DLL Side-LoadingWavesSvc64.exe + DuiLib_u.dll in Temp
Defense EvasionT1562.001 – Impair DefensesHigh-integrity PowerShell Defender exclusion
PersistenceT1053.005 – Scheduled TaskBatteries.job at system startup
PersistenceT1547.001 – Registry Run KeyHKCU\SOFTWARE\IpDates_info C2 config
C2T1095 – Non-Standard ProtocolCustom binary over TCP/443 (not HTTPS)
Defense EvasionT1070.004 – File DeletionPeriodic core file deletion + rewrite

Detection Opportunities

# WavesSvc64.exe spawned from non-Waves parent (abnormal)
process.name = "WavesSvc64.exe" AND
NOT (parent.name IN ("services.exe", "svchost.exe") OR
     file.path CONTAINS "Waves")

# rundll32.exe launched with no DLL,Export arguments
process.name = "rundll32.exe" AND
(CommandLine = "rundll32.exe" OR CommandLine = "rundll32")
# → ValleyRAT WinosStager plugin loading pattern

# Scheduled task named "Batteries" created by non-system process
EventID = 4698 AND
TaskName CONTAINS "Batteries" AND
NOT process.name IN ("svchost.exe", "taskeng.exe")

# IpDates_info registry key creation
TargetObject CONTAINS "IpDates_info"

# Defender exclusion from high-integrity non-system process
process.name IN ("powershell.exe") AND
CommandLine CONTAINS "Add-MpPreference" AND
IntegrityLevel = "High" AND
NOT (parent.name IN ("msiexec.exe", "setup.exe") AND
     CodeSignature.Trusted = true)

# Heuristic: download ZIP from Cloudflare R2 → immediate NSIS extract → Waves audio binary
# Correlate: browser download → temp zip → WavesSvc64.exe in Temp within 60s

Log Sources

  • Sysmon Event 1 (process creation with command line)
  • Sysmon Event 12/13 (registry set — IpDates_info)
  • Sysmon Event 11 (file creation in %TEMP%)
  • Windows Security 4698 (scheduled task created)
  • Network: DNS for yandibaiji0203[.]com; TCP to 161.248.87.250:443 with non-TLS ClientHello

Detection Coverage

SourceStatus
Sigma✅ DLL sideloading: image_load_side_load_from_non_system_location.yml
Sigma✅ Defender exclusion: multiple existing rules
Sigma✅ rundll32 no args: proc_creation_win_susp_bad_opsec_sacrificial_processes.yml
Sigma❌ Gap: ValleyRAT / Winos4.0 specific indicators
Sigma❌ Gap: Batteries.job scheduled task name
Sigma❌ Gap: IpDates_info registry key
Sigma❌ Gap: WavesSvc64.exe execution from non-Waves path
Splunk ESCU❌ Gap
Elastic❌ Gap

High-value rule to write: Sigma for rundll32.exe with no arguments (ValleyRAT WinosStager injection pattern). Also: scheduled task name blocklist including “Batteries.”

Sources


4. APT28 Operation MacroMaze: Webhook URLs as Office Macro C2

What’s New

S2 Grupo’s LAB52 published an analysis of APT28’s “Operation MacroMaze” — a September 2025–January 2026 campaign targeting Western and Central European organizations using Office macros that POST data to legitimate webhook service URLs (Discord, Teams, workflow automation platforms) instead of attacker-controlled IPs. This bypasses IP reputation firewalls and forces defenders to look for behavioral signals: Office applications making outbound HTTP POST requests to webhook infrastructure.

Technical Details

FieldValue
AttributionAPT28 (Fancy Bear / FANCY BEAR / Sofacy)
Campaign nameOperation MacroMaze
Activity periodSeptember 2025 – January 2026
TargetsWestern and Central European entities
Initial vectorSpear phishing with weaponized Office docs (policy docs, invoices)
Macro actionVBA macro constructs HTTP POST to webhook URL
C2 channelLegitimate webhook services (Discord webhooks, Teams incoming webhooks, automation platforms)
Data exfiltratedSystem info, credentials — encoded in webhook payload body
Defense bypassTraffic to legitimate domain IPs; no known-bad destination
ToolingStandard Windows utilities + scripting — low footprint

TTPs

TacticTechniqueObservable
Initial AccessT1566.001 – Spearphishing AttachmentOffice doc with VBA macro
ExecutionT1204.002 – User ExecutionUser enables macro / MOTW bypassed
ExecutionT1059.005 – Visual BasicVBA macro constructs HTTP POST
CollectionT1119 – Automated CollectionSystem info, credential harvest
ExfiltrationT1041 – Exfiltration Over C2HTTP POST to webhook URL (encoded body)
C2T1102.002 – Bidirectional Communication via Web ServiceWebhook receive instructions + exfil

Detection Opportunities

# Office process making outbound HTTP/HTTPS POST to non-Microsoft domains
process.name IN ("WINWORD.EXE", "EXCEL.EXE", "POWERPNT.EXE") AND
event.type = "network_connection" AND
DestinationPort IN (80, 443) AND
NOT (DestinationHostname ENDSWITH ".microsoft.com" OR
     DestinationHostname ENDSWITH ".office.com" OR
     DestinationHostname ENDSWITH ".office365.com")

# Known webhook provider domains accessed from Office apps
DestinationHostname IN (
  "discord.com", "discordapp.com",
  "webhook.site", "hooks.zapier.com",
  "webhook.office.com",  ← Note: legitimate but abused
  "pipedream.net", "make.com"
) AND
process.name IN ("WINWORD.EXE", "EXCEL.EXE", "POWERPNT.EXE")

# VBA spawning child processes (ASR: Block Office apps from creating child processes)
parent.name IN ("WINWORD.EXE", "EXCEL.EXE", "POWERPNT.EXE") AND
child.name IN ("powershell.exe", "cmd.exe", "wscript.exe", "cscript.exe")

# Encoded POST body heuristic
# Hunt for Base64-encoded strings in HTTP POST bodies from Office processes
# via proxy/web gateway inspection

KQL (Sentinel/Defender):

DeviceNetworkEvents
| where InitiatingProcessFileName in~ ("WINWORD.EXE", "EXCEL.EXE", "POWERPNT.EXE")
| where RemotePort in (80, 443)
| where RemoteUrl !contains "microsoft" and RemoteUrl !contains "office365"
| where RemoteUrl !contains "office.com"
| project Timestamp, DeviceName, InitiatingProcessFileName, RemoteUrl, RemoteIP

Log Sources

  • Office 365 audit logs (macro execution events)
  • Proxy/web gateway (HTTP POST from Office processes)
  • Sysmon Event 3 (network connection from Office binaries)
  • Windows Security Event 4688 (child processes spawned by Office)
  • Defender ASR rule telemetry

Detection Coverage

SourceStatus
Sigma✅ Partial — Office spawning child processes (multiple rules)
Sigma❌ Gap: Webhook domain outbound from Office process (specific webhook providers)
Splunk ESCU✅ Partial — Office macro network connection detection
Elastic❌ Gap: Webhook-specific Office C2 pattern

Write this: Sigma rule for outbound connections to known webhook domains (Discord, webhook.site, Zapier) from Office application processes. This is a high-fidelity detection for this TTP.

Sources


5. Seven MCP Server RCEs: eval()/exec() Epidemic in AI Tool Integrations

What’s New

Seven CVEs were published in February 2026 for MCP (Model Context Protocol) servers — all with the same root cause: user-controlled input reaching eval(), exec(), or equivalent dangerous functions without sanitization. Most servers are unauthenticated and network-exposed. The researcher (kai_security_ai) maintains a dataset of 554 network-exposed MCP servers, of which 37% have no authentication. This is the beginning of a disclosure wave, not the peak.

Technical Details

CVEServerVulnerable FunctionRoot Cause
CVE-2026-0755gemini-mcp-toolexecAsyncShell metacharacter injection
CVE-2026-1977mcp-vegalite-servereval(vegalite_specification)Python eval on user input
CVE-2026-2008fermat-mcpeqn_chart equationsPython eval on equation strings
CVE-2026-2131HarmonyOS-mcp-serverinput_textOS command injection
CVE-2026-2178xcode-mcp-serverrun_lldbCommand injection in debugger
CVE-2026-25546godot-mcpexecuteOperation projectPathexec() with unsanitized path
CVE-2026-26029sf-mcp-serverchild_process.execShell injection in Salesforce MCP

Attack chain: Prompt injection → MCP tool call → RCE on server host
Auth status: All unauthenticated or low-privilege at time of disclosure
Patch status: Most without patches at time of disclosure
Dataset: 554 network-exposed MCP servers; 37% no auth; ~15% have shell/exec tools

TTPs

TacticTechniqueObservable
Initial AccessT1190 – Exploit Public-Facing AppUnauthenticated MCP tool call
ExecutionT1059 – Command and ScriptingOS command via eval/exec injection
CollectionT1119 – Automated CollectionAI agent data via compromised MCP server
Lateral MovementT1021 – Remote ServicesMCP server as pivot point in agent environment

Detection Opportunities

# Network: MCP server processes making unexpected outbound connections
# (MCP servers should only respond, not initiate)
process.name IN ("node", "python", "python3") AND
event.type = "outbound_network" AND
DestinationPort NOT IN (443, 80) AND
NOT DestinationHostname IN (known_mcp_endpoints)

# Shell metacharacter patterns in MCP tool arguments
# Look for: ; | & ` $() in tool call parameters logged by MCP framework
# This requires MCP server-level audit logging (currently rare)

# Behavioral: unexpected child process from MCP server process
parent.name IN ("node", "python3") AND
parent.cmdline CONTAINS ("mcp", "server") AND
child.name IN ("sh", "bash", "cmd.exe", "powershell.exe", "curl", "wget")

# File system: unexpected writes from AI agent process
process.name IN ("node", "python3") AND
event.type = "file_create" AND
file.path NOT IN (mcp_server_working_dirs)

Audit step: Run the following to enumerate MCP servers in your environment:

# Find network-exposed MCP servers
ss -tlnp | grep -E ':(3000|8080|8000|9000)' | grep node
ps aux | grep -i "mcp\|model.context"

# Check for no-auth MCP servers
curl -s http://localhost:3000/mcp/health 2>/dev/null

Log Sources

  • Process creation logs (parent-child chains from node/python)
  • Network connection logs (unexpected egress from MCP server processes)
  • MCP server audit logs (if configured — most aren’t)
  • Application logs for eval()/exec() call patterns

Detection Coverage

SourceStatus
Sigma❌ Gap — No MCP-specific detection rules exist
Splunk ESCU❌ Gap
Elastic❌ Gap
Elastic✅ Partial — defense_evasion_genai_process_compiling_executables.toml covers related GenAI abuse

Critical gap: Zero detection coverage for MCP server exploitation. First priority: enumerate all MCP servers running in your environment, then build behavioral rules for unexpected child processes from node/python MCP processes.

Sources


6. Dragos 2026 OT/ICS Year in Review: 3 New Threat Groups, Industrial Ransomware +49%

What’s New

Dragos released its 9th Annual OT Cybersecurity Year in Review, introducing three new threat groups (AZURITE, PYROXENE, SYLVANITE), documenting KAMACITE expanding U.S. reconnaissance operations, and confirming 3,300 industrial organizations hit by ransomware in 2025 — a 49% year-over-year increase. ELECTRUM struck Polish wind/solar in December 2025. PYROXENE deployed wiper malware against Israeli critical infrastructure during regional conflict.

Technical Details

FieldValue
Ransomware groups targeting OT119 (up from 80 in 2024)
Industrial org victims3,300 (up from 1,693 in 2024)
Top targeted sectorManufacturing (>2/3 of victims)
New threat groupsAZURITE, PYROXENE, SYLVANITE
KAMACITEU.S. ICS reconnaissance (HMIs, VFDs, meters) March–July 2025
ELECTRUMPolish distributed energy attack (wind/solar) December 2025
PYROXENEMulti-year supply chain via fake LinkedIn; wiper deployed June 2025 vs. Israel
SYLVANITEInitial access provider; hands access to VOLTZITE for Stage 2
VOLTZITEStage 2 — OT workstation compromise, config file exfil
AZURITEEngineering workstation attacks; weaponizes public PoC code
Vuln requiring immediate action3% (“Now”)
Vuln addressable via compensating controls71% (“Next”)
OT vulns requiring no remediation27% (“Never”)

New Threat Groups

GroupFocusKey TTP
AZURITEEngineering workstationsRapid PoC weaponization; exfils alarm data + configs
PYROXENESupply chain → OT (IRGC-affiliated)Fake LinkedIn recruiters targeting OT operators; wiper activation
SYLVANITEInitial access providerEdge device vuln exploitation; hands off to VOLTZITE

TTPs

TacticTechniqueObservable
ReconnaissanceT1591 – Gather Victim Org InfoFake LinkedIn profiles targeting ICS operators
Initial AccessT1190 – Exploit Public-Facing AppEdge device exploitation (SYLVANITE)
CollectionT1005 – Data from Local SystemEngineering workstation config + alarm data exfil
ImpactT1485 – Data DestructionCustom wiper malware (PYROXENE, BAUXITE)
ImpactT1486 – Data Encrypted for ImpactESXi ransomware affecting OT-hosted SCADA

Detection Opportunities

# Misclassification prevention: detect Windows Server running SCADA software
# (these get hit as "IT incidents" but are OT)
# Look for: historian software, SCADA HMI processes on "IT" servers
process.name IN ("iFIX.exe", "Wonderware*.exe", "RSLinx*.exe", "FactoryTalk*.exe",
                 "Ignition-gateway.exe", "wincc*.exe") AND
host.role = "server"

# AZURITE pattern: rapid PoC code execution on engineering workstations
# Alert on public exploit frameworks on HMI/EWS hosts
process.name IN ("python.exe", "python3", "ruby", "go") AND
CommandLine CONTAINS ("exploit", "payload", "shellcode") AND
host.type IN ("engineering_workstation", "HMI")

# Ransomware targeting ESXi hosting OT apps
# Look for ESXi encryption of datastores containing OT VM names
vm.name MATCHES ("SCADA", "Historian", "HMI", "PLC_Gateway", "DCS") AND
event.type = "file_encrypt"

# LinkedIn phishing targeting OT operators (email gateway)
# Sender domain pattern: LinkedIn typosquats, recruitment lures
email.subject MATCHES ("opportunity", "position", "role", "engineer") AND
attachment.type IN (".docx", ".pdf") AND
recipient.department IN ("operations", "control_systems", "engineering")

Log Sources

  • ICS/OT endpoint logs (Windows on engineering workstations, HMIs)
  • ESXi hypervisor logs (datastore access, VM encryption events)
  • Email gateway (LinkedIn-themed phishing detection)
  • Network: OT zone segmentation logs, inter-zone traffic

Detection Coverage

SourceStatus
Sigma❌ Gap — No rules for KAMACITE/ELECTRUM/PYROXENE/AZURITE/SYLVANITE
Sigma❌ Gap — No engineering workstation SCADA process anomaly rules
Splunk ESCU❌ Gap for new groups
Elastic❌ Gap

Key insight: Most industrial ransomware incidents are being mislabeled “IT incidents” when ESXi hosts running SCADA/historian software are encrypted. Ensure your detection taxonomy includes SCADA application presence as an OT severity amplifier.

Sources


7. DFIR Report: Apache ActiveMQ CVE-2023-46604 → LockBit (90-Minute Detection Window)

What’s New

The DFIR Report published a detailed intrusion case study of a February 2024 incident where CVE-2023-46604 (Apache ActiveMQ ClassInfo RCE) was exploited twice on the same unpatched server — 18 days apart — ultimately delivering LockBit ransomware via RDP. Critical finding for detection engineers: if an org only detected the second intrusion, they had under 90 minutes before ransomware execution. The intrusion chain (Metasploit → GetSystem → LSASS → lateral movement → AnyDesk → LockBit) is fully documented with timing.

Technical Details

FieldValue
CVECVE-2023-46604 (Apache ActiveMQ ClassInfo RCE)
CVSS10.0
Exploit methodJava Spring class + malicious XML via CertUtil download
Initial payloadMetasploit stager
Post-exploit 1GetSystem → LSASS dump → lateral movement
Re-access (18 days later)Same unpatched server, same C2
Time to ransomware (TTR)419 hours (from first access)
TTR (if second intrusion only)< 90 minutes
Lateral movementRemote service execution + RDP
Exfil + persistenceAnyDesk installed on beachhead
DiscoveryAdvanced IP Scanner (renamed binary)
Ransomware deliveryLockBit dropped to Downloads folder + double-click execution
Ransomware variantLockBit builder (modified ransom note, Session messaging for C2)

Intrusion Timeline

Day 0: CVE-2023-46604 exploit → CertUtil download → Metasploit stager
Day 0 +40min: GetSystem → LSASS access → SMB scanning → lateral movement
Day 1: Discovery commands → threat actor loses access
Day 18: Same exploit → re-enter same server → LSASS → domain admin check
Day 18 +20min: Lateral movement to DCs via remote services
Day 18 +20min: RDP to backup server, file server
Day 18 +60min: AnyDesk install, renamed IP Scanner, LockBit dropped
Day 18 +120min: LockBit execution across environment via RDP

TTPs

TacticTechniqueObservable
Initial AccessT1190 – Exploit Public-Facing AppCVE-2023-46604 Java deserialization
ExecutionT1059 – Command & ScriptingCertUtil downloading payload from remote server
Privilege EscalationT1068 – Exploit VulnGetSystem via Metasploit
Credential AccessT1003.001 – LSASS MemoryLSASS accessed by Metasploit process
DiscoveryT1046 – Network Service ScanRenamed Advanced IP Scanner
Lateral MovementT1021.001 – RDPRDP with harvested domain admin creds
Lateral MovementT1569.002 – Service ExecutionRemote service for Metasploit payload
PersistenceT1219 – Remote Access ToolsAnyDesk installed on beachhead
ImpactT1486 – Data EncryptedLockBit deployed via double-click on RDP session

Detection Opportunities

# CVE-2023-46604: CertUtil downloading Java/XML payload
process.name = "certutil.exe" AND
CommandLine CONTAINS ("-urlcache" OR "-f") AND
(CommandLine CONTAINS ".class" OR CommandLine CONTAINS ".xml") AND
parent.name = "java.exe"

# Metasploit GetSystem patterns
# Multiple privilege escalation techniques; look for:
process.name IN ("cmd.exe", "powershell.exe") AND
parent.integrity_level = "Medium" AND
child.integrity_level = "System" AND
CommandLine CONTAINS "whoami /priv"

# LSASS access by non-credential manager processes
event.type = "process_access" AND
target.name = "lsass.exe" AND
NOT process.name IN ("lsass.exe", "svchost.exe", "csrss.exe",
                     "wininit.exe", "winlogon.exe", "taskmgr.exe")

# Renamed Advanced IP Scanner (discovery tool masquerade)
# Hash check or PE export name vs filename mismatch
OriginalFilename = "Advanced IP Scanner" AND
process.name NOT IN ("Advanced_IP_Scanner.exe", "advanced_ip_scanner.exe")

# AnyDesk installation from unusual parent
process.name = "anydesk.exe" AND
parent.name NOT IN ("explorer.exe", "msiexec.exe") AND
event.type = "process_create"

# LockBit dropped to user Downloads
file.path CONTAINS "\\Downloads\\" AND
file.extension IN (".exe") AND
sha256 NOT IN (known_clean_hashes) AND
process.name = "explorer.exe"  # double-click execution

Log Sources

  • Sysmon Events 1 (process create + command line), 10 (process access — LSASS)
  • Windows Security Event 4688 (process creation)
  • ActiveMQ logs (Java exception traces indicating ClassInfo deserialization)
  • Network: Java process outbound connections; CertUtil outbound HTTP
  • RDP logs (Event 4624 Type 10 — remote logon)

Detection Coverage

SourceStatus
Sigma✅ LSASS access: multiple existing rules
Sigma✅ CertUtil download: proc_creation_win_certutil_download.yml
Sigma✅ PsExec lateral movement: 36 rules
Sigma❌ Gap: CVE-2023-46604 Java ClassInfo exploit pattern
Sigma❌ Gap: ActiveMQ-specific exploit chain
Sigma✅ AnyDesk: existing remote tool detection rules
Splunk ESCU✅ LSASS, CertUtil detections exist
Elastic✅ LSASS, remote tool, credential dumping rules
Elastic❌ Gap: ActiveMQ-specific exploit detection

Key takeaway: If your org still has internet-exposed Apache ActiveMQ instances, patch CVE-2023-46604 immediately — it remains actively weaponized. The 90-minute window for detection on re-entry makes LSASS alerting and lateral movement detection non-negotiable.

Sources


Detection Priority Summary

PriorityThreatKey Gap
🔴 CriticalClickFix + Matanbuchus 3.0 + AstarionRATUSOShared staging, mixed-case msiexec, AstarionRAT C2
🔴 CriticalSeven MCP Server RCEsZero coverage — audit your MCP attack surface NOW
🔴 CriticalSilver Fox / ValleyRATrundll32 no-args, Batteries task, IpDates_info key
🟠 HighAPT28 Operation MacroMazeWebhook domains from Office process (write this rule)
🟠 HighActiveMQ → LockBitActiveMQ ClassInfo exploit chain (still in wild)
🟡 MediumDragos OT ReportOT mislabeling; AZURITE/PYROXENE no coverage
🟢 IntelCrowdStrike 2026 GTRReview AI platform audit logging gaps

Coverage Index Updates

Topics added today:

  • CrowdStrike 2026 Global Threat Report (Feb 24)
  • Silver Fox APT / Fake Huorong / ValleyRAT (Feb 24)
  • APT28 Operation MacroMaze (Feb 24)
  • Seven MCP Server RCEs / eval() epidemic (Feb 24)
  • ClickFix + Matanbuchus 3.0 + AstarionRAT (Feb 23)
  • Dragos 2026 OT/ICS Year in Review / AZURITE/PYROXENE/SYLVANITE (Feb 23)
  • DFIR Report: Apache ActiveMQ CVE-2023-46604 → LockBit (Feb 23)