Detection Engineering  ·  SOC Operations

SIEM & Detection
Engineering Hub

Your complete Detection Engineering reference • 226 use cases with KQL, SPL & Sigma, SOC maturity KPIs, MITRE ATT&CK coverage analysis, log volume estimation, platform comparison, and India-specific compliance mapping. Everything a detection engineer needs in one place.

226
Use Cases
5
Platforms
14
MITRE Tactics
India
Reg. Mapped
LIVE SOC METRICS BENCHMARK
Industry Avg 2025
MTTD (Mean Time to Detect)194 hrs
MTTR (Mean Time to Respond)16 days
False Positive Rate (avg)45%
Alert-to-Incident Ratio1,000:1
MITRE ATT&CK Coverage (avg)22%
Log Sources Covered (avg)14
SOC Analyst Burnout Rate65%
Tier 1 Alert Automation Rate28%
📊

SIEM & Detection Engineering Hub

Detection Engineering is the discipline of building, testing, and maintaining detection logic that finds real attacks in the noise. This hub covers everything from writing your first KQL query to building a mature detection programme with SLAs, coverage metrics, and continuous improvement pipelines.

KQLSPLSigmaMITRE ATT&CKCERT-In
📚

Detection Use Case Library

226 production-ready detection use cases covering all 14 MITRE ATT&CK tactics. Each use case includes the threat scenario, log sources required, detection logic in KQL/SPL, false positive considerations, and tuning guidance.

226 Use Cases14 TacticsKQL + SPL
ℹ️ Detection Engineering Principle
"A detection rule is only as good as its last test. Every rule must have a defined true-positive scenario, a false-positive rate target, and a tuning SLA. If you can't describe what should trigger it, you shouldn't deploy it."
Top-Priority Use Cases • India APT Targeting
Use CaseTacticTechniqueSeverityLog SourcesPlatform
Suspicious PowerShell Encoded CommandExecutionT1059.001CriticalWindows Event 4104, Sysmon 1KQL / SPL
LSASS Memory Dump (Credential Theft)Credential AccessT1003.001CriticalSysmon 10, Windows 4656KQL / SPL
Kerberoasting Attack DetectionCredential AccessT1558.003CriticalWindows 4769 (AD)KQL / SPL
Pass-the-Hash / Pass-the-TicketLateral MovementT1550.002CriticalWindows 4768/4769/4624KQL / SPL
DCSync Attack (Domain Replication)Credential AccessT1003.006CriticalWindows 4662 (AD DS)KQL / SPL
SideWinder Spearphishing DocumentInitial AccessT1566.001CriticalEmail Gateway, EDRKQL / SPL
Ransomware Shadow Copy DeletionImpactT1490CriticalWindows Event 524/4688KQL / SPL
CRIMSON RAT (APT36) C2 BeaconingC&CT1071.001CriticalProxy, DNS, FirewallKQL / SPL
Suspicious Scheduled Task CreationPersistenceT1053.005HighWindows 4698/4702, Sysmon 1KQL / SPL
Living-off-the-Land (LOLBin) AbuseDefense EvasionT1218HighSysmon 1/7, Windows 4688KQL / SPL
Cloud Storage Exfiltration (AWS S3)ExfiltrationT1537HighAWS CloudTrail, S3 Access LogsKQL
MFA Fatigue / Push Spam AttackCredential AccessT1621HighAzure AD Sign-in LogsKQL
DNS Tunnelling for Data ExfilExfiltrationT1048.003HighDNS Logs, Sysmon 22KQL / SPL
Privilege Escalation via Token ImpersonationPrivilege EscalationT1134.001HighWindows 4672/4673KQL / SPL
View all 226 use cases →
📻

Detection Query Reference

Production-ready detection queries in KQL (Microsoft Sentinel), SPL (Splunk), and XQL (Cortex XSIAM). Each query includes the threat scenario, required log sources, tuning notes, and expected false positive rate.

KQLSPLXQLSigma
PowerShell Encoded Command Execution
KQL • Microsoft Sentinel
// PowerShell Encoded Command Detection
SecurityEvent
| where EventID == 4104
| where ScriptBlockText has_any ("-EncodedCommand", "-enc", "-ec")
| where ScriptBlockText !has "Microsoft.PowerShell"
| project TimeGenerated, Account, Computer, ScriptBlockText
| extend DecodedHint = base64_decodestring(extract(@"-[Ee][Nn][Cc].*?\s+([A-Za-z0-9+/=]+)", 1, ScriptBlockText))
| where isnotempty(DecodedHint)
SPL • Splunk
index=windows source="WinEventLog:Microsoft-Windows-PowerShell/Operational"
EventCode=4104 ScriptBlockText IN ("*-EncodedCommand*", "*-enc *", "*-ec *")
NOT ScriptBlockText="*Microsoft.PowerShell*"
| eval decoded=base64decode(mvindex(split(ScriptBlockText,"-EncodedCommand "),1))
| table _time, ComputerName, UserID, ScriptBlockText, decoded
LSASS Memory Access (Credential Theft)
KQL • Microsoft Sentinel (Sysmon)
// Detects process access to LSASS • Mimikatz, Cobalt Strike, etc.
Event
| where Source == "Microsoft-Windows-Sysmon"
| where EventID == 10
| parse EventData with * '<TargetImage>' TargetImage '</TargetImage>' *
| where TargetImage has "lsass.exe"
| parse EventData with * '<SourceImage>' SourceImage '</SourceImage>' *
| where SourceImage !has_any ("svchost.exe", "MsMpEng.exe", "csrss.exe")
| project TimeGenerated, Computer, SourceImage, TargetImage
Kerberoasting Detection
KQL • Microsoft Sentinel (Windows Security)
// Kerberoasting: TGS requests for RC4 encryption (0x17 = Etype 23)
SecurityEvent
| where EventID == 4769
| where TicketEncryptionType == "0x17"
| where ServiceName !startswith "$"
| where ServiceName !in ("krbtgt", "kadmin")
| summarize RequestCount=count(), Accounts=make_set(Account) by bin(TimeGenerated, 1h), Computer, ServiceName
| where RequestCount > 5
| order by RequestCount desc
Ransomware • Shadow Copy Deletion
KQL • Microsoft Sentinel
// Ransomware pre-encryption: deleting shadow copies to prevent recovery
SecurityEvent
| where EventID == 4688
| where NewProcessName has_any ("vssadmin.exe","wmic.exe","bcdedit.exe","wbadmin.exe")
| where CommandLine has_any ("delete shadows","resize shadowstorage","recoveryenabled No","delete systemstatebackup")
| project TimeGenerated, Account, Computer, CommandLine, NewProcessName
| extend Severity = "CRITICAL • Ransomware Stage 4 Indicator"
Σ

Sigma Rules • Platform-Agnostic Detection

Sigma is the vendor-agnostic detection rule format • write once, convert to KQL, SPL, XQL, or Elasticsearch DSL. The community repository has 3,000+ rules. This section covers Sigma fundamentals, structure, and top rules for India-targeting threats.

YAML Format3000+ Community Rulessigmac / pySigma
Sigma Rule Structure
Sigma YAML • PowerShell Encoded Command
title: PowerShell Encoded Command Execution
id: 8d72c5bb-f461-4c91-a5f8-a6c81fa70854
status: production
description: Detects PowerShell execution with encoded command to bypass detection
references:
- https://attack.mitre.org/techniques/T1059/001/
author: Cyber Handbook
date: 2025/01/15
tags:
- attack.execution
- attack.t1059.001
logsource:
product: windows
category: ps_script
detection:
selection:
ScriptBlockText|contains:
- '-EncodedCommand'
- '-enc '
- '-ec '
filter:
ScriptBlockText|contains: 'Microsoft.PowerShell'
condition: selection and not filter
falsepositives:
- Legitimate admin scripts using encoded commands
level: high
Convert Sigma to Platform Queries
pySigma • Convert to KQL
pip install pySigma-backend-microsoft365defender
sigma convert -t microsoft365defender -p windows rules/powershell_encoded.yml
sigmac • Convert to SPL
pip install sigmatools
sigmac -t splunk -c splunk-windows rules/powershell_encoded.yml
Browse 3,000+ Sigma Rules →
📊

SOC KPIs & Performance Metrics

A high-performing SOC measures everything that matters and ignores vanity metrics. These are the KPIs that Gartner, MITRE, and leading security operations programmes use to measure detection effectiveness, analyst efficiency, and programme maturity. Track these monthly and present to the board quarterly.

Gartner SOC FrameworkNIST CSFMITRE D3FEND
⚠️ Metric Trap • Avoid These
Alert volume is NOT a KPI • high alert counts indicate tuning failure. Tickets closed is NOT a KPI • it rewards false-positive closure. "Incidents handled" is NOT a KPI without severity weighting. Measure outcomes, not activity.
Tier 1 • Detection Effectiveness KPIs
MTTD
< 1 hour
Mean Time to Detect • from attacker first action to first SIEM alert. Industry average: 194 hours. World-class: <1 hour.
Gartner Target: <60 min
MTTR
< 4 hours
Mean Time to Respond • from first alert to containment action. Industry average: 16 days. World-class: <4 hours for critical.
NIST Target: <4 hours critical
False Positive Rate
< 10%
% of alerts that are not real incidents. Industry average: 45%. Every false positive costs analyst time. Target: <10% per rule.
Industry Avg: 45%
Detection Coverage
> 60%
% of MITRE ATT&CK techniques covered by at least one production rule. Most orgs: 22%. Target: >60% for mature SOC.
Industry Avg: 22%
True Positive Rate
> 90%
% of alerts that represent real malicious activity. Inverse of false positive rate. Indicates rule quality and tuning effectiveness.
Target: >90% per rule
Alert-to-Incident Ratio
< 50:1
Alerts generated per confirmed incident. Ratio of 1,000:1 is common • indicates massive tuning opportunity. Target: <50:1 with SOAR.
Industry Avg: 1000:1
Tier 2 • Analyst Efficiency KPIs
Alerts per Analyst / Day
< 20
Workload per analyst. Above 50 alerts/day leads to burnout and missed detections. SOAR automation should reduce manual handling below 20.
Burnout threshold: 50+/day
Analyst Escalation Rate
< 15%
% of T1 alerts escalated to T2. High escalation = poor playbooks or undertrained T1. Low escalation without context = missed escalations.
Target: 10•15%
Dwell Time
< 14 days
Time from initial compromise to detection. IBM Cost of Breach 2024: avg 194 days. Target: detect within 14 days for advanced threats.
IBM Avg: 194 days
Playbook Coverage
> 80%
% of alert types with documented/automated response playbooks. High playbook coverage reduces MTTR and analyst decision fatigue.
Target: >80% alert types
Automation Rate (SOAR)
> 60%
% of L1 alerts handled fully automatically without human intervention. World-class SOCs automate 80%+ of repetitive alert types.
World-class: 80%+
Mean Time to Acknowledge
< 15 min
Time from alert creation to analyst acknowledgement. SLA breach indicates understaffing or alert fatigue. Critical alerts: <5 minutes.
Critical SLA: <5 min
Tier 3 • Programme Health KPIs
New Rules Deployed / Month
≥ 5
Detection rule production velocity. A healthy detection programme continuously adds new detections. Track per analyst and per team.
Healthy: 5•20/month
Rule Deprecation Rate
≤ 10%
% of rules deprecated per quarter. High deprecation = poor initial quality. Zero deprecation = no rule lifecycle management.
Healthy: 5•10%/quarter
Log Source Gaps
0 Tier 1
Critical log sources not onboarded (Tier 1: AD, EDR, Firewall, DNS). Any Tier 1 gap is a critical detection blind spot requiring immediate action.
Tier 1 gaps = Critical risk
Purple Team Exercise Cadence
Quarterly
Frequency of detection validation exercises against real attacker TTPs. Quarterly purple team exercises validate rule effectiveness against current threats.
Minimum: Semi-annual
Threat Hunt Cadence
Monthly
Frequency of proactive threat hunting operations. Monthly hunts targeting specific hypotheses based on threat intelligence.
Minimum: Quarterly
SIEM Data Retention Compliance
180 days
CERT-In mandates 180-day log retention in India. RBI requires 1 year for financial sector. Track compliance monthly • regulatory risk if breached.
CERT-In: 180 days mandatory
⭐

SOC Maturity Model

SOC maturity is not about headcount • it's about process, automation, and measurable outcomes. This model maps 5 maturity levels across 8 capability domains. Use it to baseline your current state and build a credible 12-month improvement roadmap for CISO and board reporting.

Gartner SOC Maturity ModelCMMI Inspired8 Domains
CapabilityL1 InitialL2 DevelopingL3 DefinedL4 ManagedL5 Optimising
Alert TriageManual, ad hocBasic playbooksDocumented SLASOAR-automated 60%90%+ auto-resolved
Detection Coverage<10% ATT&CK10•25% ATT&CK25•50% ATT&CK50•75% ATT&CK>75% ATT&CK
Threat IntelligenceNo TI programmeIOC blocking onlyTIP with SIEM integrationThreat-driven detectionIntel-led hunt operations
Log Coverage<5 sources5•15 sourcesAll Tier 1 sourcesFull stack incl. cloudUniversal telemetry
Incident ResponseNo IR planBasic IR runbookTested IR playbooksSOAR-orchestrated IRAutonomous containment
Threat HuntingNo huntingAd hoc investigationsHypothesis-driven huntsMonthly structured huntsContinuous hunting pipeline
MTTD Target>30 days7•30 days1•7 days1•24 hours<1 hour
False Positive Rate>70%40•70%20•40%10•20%<10%
🔎 India SOC Maturity Reality
Based on CERT-In and industry assessments, 78% of Indian organisations operate at L1•L2. BFSI sector is most mature (avg L3). Government and healthcare are predominantly L1. Most organisations can reach L3 within 18 months with investment in SOAR, log coverage, and detection engineering skills.
📈

Log Volume & Cost Estimator

Estimate daily EPS (Events Per Second), GB/day ingest volume, and annual SIEM licensing cost based on your infrastructure. Accurate sizing prevents budget overruns • SIEM cost surprises are the #1 reason SOC programmes are scaled back.

EPS CalculatorStorage SizingCost Modelling
Infrastructure Input
Full Interactive Estimator →
🔯

MITRE ATT&CK Detection Coverage

Map your detection rules to the 14 MITRE ATT&CK tactics. Coverage analysis drives prioritisation • focus on the tactics most used by APT groups targeting India and your sector, not just the most common generic threats.

14 Tactics200+ TechniquesIndia APT Focus
TA0001
Initial Access
78% • 14/18 techniques
TA0002
Execution
72% • 10/14 techniques
TA0003
Persistence
55% • 12/22 techniques
TA0004
Priv. Escalation
60% • 9/15 techniques
TA0005
Defense Evasion
38% • 16/42 techniques
TA0006
Credential Access
82% • 13/16 techniques
TA0007
Discovery
45% • 10/23 techniques
TA0008
Lateral Movement
75% • 6/8 techniques
TA0009
Collection
35% • 7/20 techniques
TA0011
C&C
50% • 10/20 techniques
TA0010
Exfiltration
42% • 7/17 techniques
TA0040
Impact
65% • 9/14 techniques
Interactive Coverage Matrix →
🇮🇳

India SIEM Compliance Mapping

SIEM is no longer optional for Indian regulated entities. CERT-In, RBI, SEBI, and IRDAI all mandate log collection, retention, and alerting for specific event types. This mapping shows which SIEM controls satisfy which regulatory requirements.

CERT-In 2022RBI IT FrameworkSEBI CSCRF 2024DPDP Act 2023
Regulatory RequirementRegulatorSIEM ControlLog Sources RequiredRetention
6-Hour Incident ReportingCERT-InAlert on 20 incident categories, auto-notify workflowAll log sources, EDR, Firewall, Email180 days
180-Day Log Retention in IndiaCERT-InSIEM storage with India-region data residencyAll ICT system logs180 days
VPN/Remote Access LoggingCERT-InIngest and retain all VPN authentication logsVPN Gateway, IAM, PAM180 days
NTP Sync MonitoringCERT-InAlert on NTP drift >5 seconds from approved serversSyslog, NTP Logs180 days
Privileged User Activity MonitoringRBIUBA rules for admin account anomalies, PAM session alertsAD, PAM, Windows 46721 year
Cybersecurity Incident DetectionRBISIEM alerting for all RBI-defined incident categoriesFull stack1 year
SOC Operations MandateSEBI24x7 SIEM monitoring with documented escalationTrading systems, Market data5 years
Annual Cyber Audit EvidenceSEBISIEM compliance reports, alert logs, IR evidenceAll SIEM events5 years
Personal Data Breach DetectionDPDP ActDLP alerts, data access anomaly detection, exfil detectionDLP, EDR, Proxy, EmailAs needed
🖥

SIEM Platform Comparison

Choosing the wrong SIEM is a multi-year, multi-million rupee mistake. This comparison covers the 5 leading platforms on query language, pricing model, India data residency, and total cost of ownership.

Gartner MQ: SIEM 2024India Deployment
📊 Microsoft Sentinel
Query: KQL
  • Pay-per-GB • scales down at low volume
  • 200+ native connectors for M365/Azure
  • India region (Central India) available
  • SOAR via Logic Apps included
  • Free Defender data ingest for E5
  • Cost spikes at high log volume
  • KQL learning curve for SPL teams
  • Weaker on-prem log source support
Best for: Azure/M365 shops~$2.46/GB
🔍 Splunk Enterprise
Query: SPL
  • Unmatched search flexibility with SPL
  • Largest app/addon ecosystem (2000+)
  • Excellent on-prem deployment support
  • Splunk ES + SOAR (Phantom) bundleable
  • Highest TCO • ingest-based pricing expensive
  • Complex licensing negotiation
  • Cisco acquisition adds roadmap uncertainty
Best for: Large SOC, custom detection~$150/GB/yr
💥 Google Chronicle
Query: YARA-L / UDM
  • Flat-fee pricing • unlimited data ingest
  • VirusTotal + Mandiant intel built-in
  • Petabyte-scale without cost penalty
  • 1-second search across years of data
  • India data residency options limited
  • YARA-L learning curve
  • Smaller connector ecosystem
Best for: High-volume log environmentsFlat fee (negotiated)
🚀 Elastic Security
Query: EQL / KQL-like
  • Open-source core • no vendor lock-in
  • EQL (Event Query Language) for sequences
  • Self-hosted option for data sovereignty
  • Strong India deployment community
  • Free tier for small environments
  • Requires significant operational investment
  • Less mature SOAR integration
  • Enterprise support expensive
Best for: Budget-conscious, data sovereigntyFree OSS / $95/node/mo
🛡 IBM QRadar
Query: AQL
  • Strong out-of-box correlation rules
  • On-prem deployment for air-gapped environments
  • Dominant in India PSU banks & govt
  • Deep network flow analysis
  • Palo Alto asset acquisition adds uncertainty
  • UI lagging modern platforms
  • High professional services dependency
Best for: On-prem, Govt, BFSIEPS-based licensing
💾

Log Source Priority Guide

Not all log sources are equal. Tier 1 sources are critical • missing any of them creates detection blind spots that attackers reliably exploit. This priority framework is used by mature SOC programmes to sequence log onboarding.

CERT-In MandatedTier 1•4
Log SourceTierKey EventsEPS (est.)CERT-In Required
Windows Security Events (DC)Tier 14624,4625,4648,4662,4769,4768,7045500•2000✅ Yes
Sysmon (Windows Endpoints)Tier 11,3,7,10,11,13,22,25200•1000 per hostRecommended
Perimeter FirewallTier 1Allowed/denied connections, geo-blocks2000•50000✅ Yes
EDR Telemetry (CrowdStrike/Defender)Tier 1Process, network, file, registry events1000•5000 per hostRecommended
DNS LogsTier 1All DNS queries • C2 detection, DGA, tunnelling5000•100000✅ Yes
VPN / Remote AccessTier 1Auth, failures, geo-anomalies50•500✅ Yes (CERT-In mandated)
Azure AD / Entra IDTier 1Sign-ins, MFA, Conditional Access, PIM200•2000Recommended
Email Gateway (Proofpoint/Defender)Tier 2Blocked/allowed, phishing, malware100•500Recommended
Web Proxy / SWGTier 2URL categories, blocked requests, data exfil1000•10000Recommended
AWS CloudTrailTier 2API calls, IAM changes, S3 access100•5000Required if AWS used
PAM (CyberArk/BeyondTrust)Tier 2Privileged sessions, password checks, vault access50•200Required if PAM deployed
Linux Syslog / AuditdTier 2Auth, sudo, process execution, file access200•2000 per hostRecommended
🎲

Alert Tuning Playbook

Alert fatigue is the #1 SOC killer. A single poorly-tuned rule can generate 10,000 false positives per day, burying real alerts. This playbook provides a systematic approach to reducing false positives without increasing missed detections.

False Positive ReductionSOAR IntegrationRule Lifecycle
👉 The Golden Rule
Never delete an alert rule • tune it. A rule that generates 100% false positives still has value if it correctly describes a malicious pattern. Reduce scope with exceptions, not deletion. Document every tuning decision with date, reason, and reviewer.
5-Step Tuning Process
1
Measure • Run rule in observe-only mode for 7 days
Count total alerts, true positives, false positives. Calculate FP rate. If FP >50%, do NOT promote to production • tune first. If FP <10%, promote with monitoring.
2
Categorise • Split FPs into types
Type A: Legitimate activity that looks like attack (add to allowlist). Type B: Misconfigured threshold (adjust sensitivity). Type C: Missing context (add enrichment). Type D: Wrong log source (fix parser). Each type needs a different fix.
3
Enrich • Add context before alerting
Asset criticality (High/Med/Low), user risk score, geo-location, known-good baseline. An alert for a Tier 1 asset should be Critical; same alert for a test machine can be Low. Context-driven severity reduces analyst time per alert by 40%.
4
Automate • SOAR playbook for low-risk FP patterns
If SOAR can determine alert is FP in <30 seconds, automate the close. Human review for close decisions that require context outside SIEM. Track automation accuracy • if auto-close is wrong >5% of time, send back to step 2.
5
Review • Monthly rule performance report
Run monthly: rules with 0 true positives in 90 days (retire or retune). Rules with FP >30% (mandatory retune). Rules that haven't been reviewed in 6 months (flag for owner review). Detection engineering is maintenance, not deployment.