본문으로 건너뛰기
cccskills
무료GitHub에 공개

detecting-evasion-techniques-in-endpoint-logs

Examine Windows endpoint logs for attempts to avoid security detection, including log clearing, altered file timestamps, injected code, and disabled protection. Correlate events to investigate suspicious behavior and develop detection rules with context for legitimate activity.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

원문원문 설명 보기

Detects defense evasion techniques used by adversaries in endpoint logs including log tampering, timestomping, process injection, and security tool disabling. Use when investigating suspicious endpoint behavior, building detection rules for evasion tactics, or conducting threat hunting for stealthy adversary activity. Activates for requests involving evasion detection, defense evasion analysis, log tampering detection, or MITRE ATT&CK TA0005.

설치 방법 보기

이럴 때 유용합니다

  • Investigating cleared logs and altered file timestamps
  • Checking suspected attempts to disable security tools
  • Finding abuse of legitimate Windows utilities
  • Building detection rules that correlate multiple warning signs

포함된 파일(8)

  • SKILL.md9.5 KB
  • assets/template.md833 B
  • LICENSE11.0 KB
  • references/api-reference.md1.7 KB
  • references/standards.md1.6 KB
  • references/workflows.md2.2 KB
  • scripts/agent.py5.9 KB
  • scripts/process.py7.1 KB

SKILL.md(원문)

설치하기 전에 에이전트에게 전달되는 지시 내용을 확인할 수 있습니다.

Detecting Evasion Techniques in Endpoint Logs

When to Use

Use this skill when:

  • Hunting for adversary defense evasion techniques (MITRE ATT&CK TA0005) in endpoint telemetry
  • Building detection rules for common evasion methods (process injection, timestomping, log clearing)
  • Investigating incidents where adversaries disabled or bypassed security tools
  • Analyzing endpoint logs for indicators of living-off-the-land binary (LOLBin) abuse

Do not use this skill for network-level evasion (use network traffic analysis) or for malware reverse engineering.

Prerequisites

  • Sysmon installed and configured with comprehensive logging rules (SwiftOnSecurity or Olaf Hartong config)
  • Windows Security Event Log with advanced audit policy enabled
  • EDR telemetry (CrowdStrike, SentinelOne, Microsoft Defender for Endpoint)
  • SIEM platform for log correlation (Splunk, Elastic, Sentinel)
  • MITRE ATT&CK Enterprise matrix for technique reference

Workflow

Step 1: Detect Log Tampering (T1070)

Windows Event Log clearing (T1070.001):

# Sysmon Event ID 1 (Process Create) for wevtutil
EventID: 1
CommandLine contains: "wevtutil cl" OR "wevtutil clear-log"

# Security Event ID 1102 - Audit log was cleared
EventID: 1102
Source: Microsoft-Windows-Eventlog

# System Event ID 104 - Event log was cleared
EventID: 104

# PowerShell log clearing
EventID: 1 (Sysmon)
CommandLine contains: "Clear-EventLog" OR "Remove-EventLog"

# Splunk query:
index=windows (EventCode=1102 OR EventCode=104)
  OR (EventCode=1 CommandLine="*wevtutil*cl*")
  OR (EventCode=1 CommandLine="*Clear-EventLog*")
| table _time host user CommandLine EventCode

Timestomping (T1070.006):

# Sysmon Event ID 2 - File creation time changed
EventID: 2
# Look for creation times set far in the past on recently-written files
# Correlate with Event ID 11 (FileCreate) - if FileCreate is recent but
# creation time in Event ID 2 is old, timestomping is likely

# MDE Advanced Hunting (KQL):
DeviceFileEvents
| where ActionType == "FileTimestampModified"
| where Timestamp > ago(7d)
| extend TimeDiff = datetime_diff('day', Timestamp, ReportedFileCreationTime)
| where TimeDiff > 30
| project Timestamp, DeviceName, FileName, FolderPath,
    ReportedFileCreationTime, InitiatingProcessFileName

Step 2: Detect Process Injection (T1055)

# Sysmon Event ID 8 - CreateRemoteThread
EventID: 8
# Alert when source process is unusual (not system processes)
# Filter out known legitimate: antivirus, debugging tools
SourceImage NOT IN ("C:\Windows\System32\csrss.exe",
                     "C:\Windows\System32\lsass.exe")

# Sysmon Event ID 10 - ProcessAccess with suspicious access masks
EventID: 10
GrantedAccess contains: "0x1F0FFF" OR "0x1FFFFF" OR "0x001F0FFF"
# PROCESS_ALL_ACCESS = 0x1F0FFF (common in injection)
# Filter legitimate: AV accessing all processes

# Sysmon Event ID 25 - Process Tampering
EventID: 25
Type: "Image is replaced"  # Process hollowing indicator

# Splunk detection:
index=sysmon EventCode=8
| where NOT match(SourceImage, "(?i)(csrss|svchost|MsMpEng|defender)")
| stats count by SourceImage TargetImage host
| where count < 5
| sort - count

Step 3: Detect Security Tool Disabling (T1562)

# Service stopped events for security services
EventID: 7045 (new service) OR 7036 (service state change)
ServiceName IN ("WinDefend", "Sense", "CrowdStrike Falcon Sensor",
                 "SentinelAgent", "csagent", "MBAMService")

# Sysmon Event ID 1 - Processes that disable Defender
CommandLine contains: "Set-MpPreference -DisableRealtimeMonitoring"
  OR "sc stop WinDefend"
  OR "sc config WinDefend start= disabled"
  OR "net stop" AND ("windefend" OR "sense" OR "csagent")

# Registry modification to disable security features
# Sysmon Event ID 13 - Registry value set
TargetObject contains: "DisableAntiSpyware"
  OR "DisableRealtimeMonitoring"
  OR "DisableBehaviorMonitoring"
Details: "DWORD (0x00000001)"

# MDE KQL:
DeviceRegistryEvents
| where RegistryValueName in ("DisableAntiSpyware", "DisableRealtimeMonitoring")
| where RegistryValueData == "1"
| project Timestamp, DeviceName, RegistryKey, InitiatingProcessFileName

Step 4: Detect Masquerading (T1036)

# Sysmon Event ID 1 - Process with legitimate name from unusual path
EventID: 1
Image contains: "svchost.exe" AND Image NOT starts with: "C:\Windows\System32\"
Image contains: "csrss.exe" AND Image NOT starts with: "C:\Windows\System32\"
Image contains: "lsass.exe" AND Image NOT starts with: "C:\Windows\System32\"

# Process name mismatch (original filename vs. current name)
# Sysmon captures OriginalFileName from PE header
EventID: 1
OriginalFileName != (parsed filename from Image path)

# Double extension files
EventID: 11 (FileCreate)
TargetFilename matches: "*\.pdf\.exe" OR "*\.doc\.exe" OR "*\.jpg\.exe"

# Splunk:
index=sysmon EventCode=1
| eval process_name=mvindex(split(Image,"\\"),-1)
| where (process_name="svchost.exe" AND NOT match(Image,"(?i)C:\\\\Windows\\\\System32"))
  OR (process_name="csrss.exe" AND NOT match(Image,"(?i)C:\\\\Windows\\\\System32"))
| table _time host Image ParentImage CommandLine User

Step 5: Detect LOLBin Abuse (T1218, T1127)

# Common LOLBin abuse patterns:

# mshta.exe executing remote content
EventID: 1
Image ends with: "mshta.exe"
CommandLine contains: "http" OR "javascript:" OR "vbscript:"

# certutil.exe downloading files
EventID: 1
Image ends with: "certutil.exe"
CommandLine contains: "-urlcache" OR "-decode" OR "-encode"

# regsvr32.exe executing scriptlets
EventID: 1
Image ends with: "regsvr32.exe"
CommandLine contains: "/s /n /u /i:" OR "scrobj.dll"

# rundll32.exe with unusual DLLs
EventID: 1
Image ends with: "rundll32.exe"
CommandLine contains: "javascript:" OR ".js" OR "http:"

# MSBuild executing inline tasks
EventID: 1
Image contains: "MSBuild.exe"
CommandLine NOT contains: ".sln" AND NOT contains: ".csproj"

Step 6: Build Detection Rule Correlation

# Combine multiple weak signals into high-confidence detection:

# Rule: Potential post-exploitation evasion chain
# Trigger when 3+ evasion techniques observed on same host within 1 hour

# Splunk correlation search:
index=sysmon host=*
| eval technique=case(
    EventCode=2, "timestomping",
    EventCode=8 AND NOT match(SourceImage,"csrss|svchost"), "process_injection",
    EventCode=1 AND match(CommandLine,"(?i)wevtutil.*cl"), "log_clearing",
    EventCode=13 AND match(TargetObject,"DisableRealtimeMonitoring"), "security_disable",
    EventCode=1 AND match(CommandLine,"(?i)(mshta|certutil.*urlcache|regsvr32.*/s.*/n)"), "lolbin_abuse",
    true(), NULL
)
| where isnotnull(technique)
| bin _time span=1h
| stats dc(technique) as technique_count values(technique) as techniques by host _time
| where technique_count >= 3
| sort - technique_count

Key Concepts

TermDefinition
Defense Evasion (TA0005)MITRE ATT&CK tactic where adversaries attempt to avoid detection during operations
Process Injection (T1055)Technique of injecting code into another process's memory space to execute in a trusted context
Timestomping (T1070.006)Modifying file timestamps to make malicious files appear old and blend with legitimate files
Masquerading (T1036)Naming malicious files or processes to match legitimate system files to avoid detection
LOLBinLiving Off the Land Binary; legitimate Windows tool repurposed by adversaries
Indicator Removal (T1070)Clearing logs, deleting files, or modifying artifacts to remove evidence of compromise

Tools & Systems

  • Sysmon: Advanced Windows system monitoring with kernel-level visibility
  • Microsoft Defender for Endpoint: EDR with advanced hunting (KQL) for evasion detection
  • CrowdStrike Falcon: IOA-based behavioral detection for evasion techniques
  • Elastic Security: SIEM with prebuilt detection rules for ATT&CK evasion techniques
  • Sigma Rules: Vendor-agnostic detection rule format with extensive evasion rule library

Common Pitfalls

  • Alert fatigue from process injection rules: Many legitimate tools (AV, accessibility) perform process injection. Maintain an allowlist of known-good source processes.
  • Missing Sysmon Event ID 8/10: Default Sysmon configurations may not capture CreateRemoteThread or ProcessAccess. Use a comprehensive Sysmon config.
  • Ignoring parent process context: A suspicious command line from cmd.exe is concerning only if the parent of cmd.exe is unusual (e.g., Excel spawning cmd.exe).
  • Not correlating across event types: Single events are often benign. Combine multiple weak signals (process creation + network connection + file creation) for high-confidence detections.

리뷰

아직 리뷰가 없습니다. 사용해 본 소감을 남겨 주세요.

댓글

로그인 없이 쓸 수 있습니다. 사용 후기, 질문, 작동하지 않은 환경 등을 편하게 남겨 주세요.

아직 댓글이 없습니다. 첫 댓글을 남겨 보세요.

보내면 바로 공개됩니다. 이메일 주소나 전화번호 같은 개인 정보는 쓰지 마세요. URL은 2개까지입니다.

0 / 1,000자

이 양식은 reCAPTCHA로 보호되며 Google 개인정보처리방침 및 서비스 약관이 적용됩니다.

같은 저장소의 스킬

Assess exposure of Windows secrets protected by DPAPI, the operating system’s data protection service, during authorized security tests. Guide examination and decryption of saved credentials and browser data using available user sessions, credentials, or backup keys.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

  • Assessing saved credential exposure in authorized tests
  • Examining collected DPAPI files offline
  • Reviewing protected browser and RDP credentials
mukul975/Anthropic-Cybersecurity-Skills3.5만2026년 8월 31일 업데이트

Assess Shadow Credentials account takeover during authorized Active Directory testing. Examine permissions that allow public key registration, validate authentication without a password reset, and document credential cleanup, detection opportunities, and access control changes needed for remediation.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

  • Assessing risky permissions identified in BloodHound
  • Validating account takeover without resetting a password
  • Generating directory change records during security exercises
mukul975/Anthropic-Cybersecurity-Skills3.5만2026년 8월 31일 업데이트

Prepare for CMMC Level 2, a cybersecurity certification for defense contractors. Define the systems handling controlled information, review security requirements, calculate the SPRS assessment score, and organize remediation plans and evidence into a readiness report.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

  • Define the assessment scope for systems handling controlled information
  • Calculate an SPRS score and prioritize unmet security requirements
  • Prepare security plans, remediation milestones, and assessment evidence
mukul975/Anthropic-Cybersecurity-Skills3.5만2026년 8월 31일 업데이트

Create disk images for investigations using dd or dcfldd on Linux. The workflow covers write protection, bit-for-bit copying, hash checks to verify integrity, and documentation of the acquisition process.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

  • Preserving a suspect drive before investigation
  • Creating evidence copies of USB devices or memory cards
  • Verifying that a disk image matches its source
mukul975/Anthropic-Cybersecurity-Skills3.5만2026년 8월 31일 업데이트

Review Active Directory permissions with ldap3, a Python library for directory queries. Identify excessive access granted to non-administrative accounts, explain potential privilege escalation paths, and produce a JSON report with affected objects and remediation recommendations.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

  • Auditing excessive Active Directory permissions
  • Investigating privilege escalation paths from ordinary accounts
  • Reporting affected objects and permission remediation recommendations
mukul975/Anthropic-Cybersecurity-Skills3.5만2026년 8월 31일 업데이트

Inspect suspicious Android APK files without running them. Analyze permissions, app components, suspicious code patterns, and network indicators, then produce a JSON risk report to support malware triage and detection rule development.

한국어 요약은 준비 중입니다. 영어 요약을 표시하고 있습니다.

  • Triage suspicious APK files without running them
  • Inspect dangerous permission combinations and suspicious API calls
  • Extract network indicators for malware detection rules
mukul975/Anthropic-Cybersecurity-Skills3.5만2026년 8월 31일 업데이트

mukul975의 스킬 모두 보기

이 스킬의 문제 신고