Join our Newsletter — 33% off our NHI Course
Home› Guides› ATT&CK Detection Strategies and Analytics Guide: Turning MITRE…
Guide Threat Intelligence & Threat Actors

ATT&CK Detection Strategies and Analytics Guide: Turning MITRE Detection Objects into Working Rules

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 27 October 2026 14 min read
On this page

Since ATT&CK v18, the detection guidance on every technique page has moved into structured objects, and detection teams that still work from the old free-text notes or from data sources are reading a model MITRE has retired. ATT&CK now describes defence through detection strategies (DET IDs), platform-specific analytics (AN IDs), data components (DC IDs) with named log sources, and mutable elements that tell you what to tune locally. This guide explains the objects field by field, shows what the v20 dataset says about the telemetry you need, and sets out a repeatable way to turn an analytic into a tested production rule and to measure the result. It is for detection engineers, SOC leads and anyone maintaining ATT&CK-based content. For scoring and heatmaps, see the companion MITRE ATT&CK Coverage Mapping Guide; this page goes one level deeper into the objects themselves.

Key takeaways

  • One strategy per technique, several analytics per strategy. In the v20 enterprise dataset each live technique with detection content has exactly one detection strategy, and each analytic covers exactly one platform.
  • Analytics are specifications, not rules. They describe behaviour, name log sources and list the parameters you must set; you still write, test and tune the query in your own SIEM or EDR.
  • Log source references are a ready-made collection plan. Aggregating them across your priority techniques tells you which event channels to collect, and the dataset shows a handful of sources carry most of the weight.
  • Mutable elements are your tuning backlog. Fields such as TimeWindow and UserContext appear on hundreds of analytics and are where false positives are won or lost.
  • Treat ATT&CK content as versioned data. Strategies are added, patched and deprecated in every release; pin the version you mapped against and review changes at each release.

What changed and when

MITRE's v18 release notes (October 2025) state that "Detections in techniques have been replaced with Detection Strategies", and describe major updates to data components and the deprecation of data sources. Later releases kept the model and added content. The v20 enterprise technique objects carry no x_mitre_detection or x_mitre_data_sources fields at all.

ReleaseEnterprise detection strategiesEnterprise analyticsData componentsNotes
v18 (October 2025)6911,739106Model introduced; Data Sources deprecated; Mobile 124 strategies, ICS 83
v20 (13 October 2026)7031,770106Six new enterprise strategies (AI-enabled exfiltration, AI result poisoning, anti-automation evasion, application-specific passwords), twelve new analytics; Mobile 124 strategies and 211 analytics; ICS 106 strategies and 302 analytics

The v20 dataset holds 705 strategy objects: 703 live plus two deprecated ones (DET0239 and DET0317) that belonged to Impair Defenses, which v19 revoked when it split Defense Evasion into Stealth (TA0005) and Defense Impairment (TA0112). Three live enterprise techniques have no strategy yet: T1698 Indirect Prompt Injection, T1699 Direct Prompt Injection and T1685.007 Impair Autonomous Defenses. The ATT&CK Stealth and Defense Impairment Tactics Guide covers the tactic split.

Anatomy of the objects

The objects are STIX 2.1 custom types defined in MITRE's ATT&CK Data Model. The table summarises the fields detection engineers use, from the published schema and the v20 data.

ObjectID formatKey fieldsWhat it means for you
Detection strategy (x-mitre-detection-strategy)DET0000name, x_mitre_analytic_refs, x_mitre_version; linked to a technique by a detects relationshipThe behavioural approach to catching one technique, across platforms
Analytic (x-mitre-analytic)AN0000description, x_mitre_platforms (exactly one), x_mitre_log_source_references, x_mitre_mutable_elementsThe platform-specific logic you translate into a query
Log source referenceInside an analyticx_mitre_data_component_ref, name (for example WinEventLog:Sysmon), channel (for example EventCode=10)The exact event stream the analytic expects
Data component (x-mitre-data-component)DC0000description, x_mitre_log_sources (name and channel pairs)A vendor-neutral category of telemetry, such as DC0032 Process Creation
Mutable elementInside an analyticfield, descriptionA parameter MITRE expects you to set for your environment

The schema requires each log source reference's name and channel to come from the referenced data component's own x_mitre_log_sources list. In the v20 enterprise data, 4,202 of 4,206 references match exactly, so you can join analytics to data components reliably. Forty-nine analytics carry no log source references, mostly for pre-compromise platforms, and should be treated as guidance rather than buildable detections.

On attack.mitre.org a strategy page shows the technique detected, ID, domains, analytic IDs, version and dates, then one pane per analytic with a Log Sources table (Data Component, Name, Channel) and a Mutable Elements table (Field, Description).

Reading a strategy: a worked example

DET0546 Detection of Abused or Compromised Cloud Accounts for Access and Persistence detects T1078.004 Cloud Accounts and has four analytics:

AnalyticPlatformBehaviourLog sources (name: channel)Mutable elements
AN1503Identity ProviderSign-ins from impossible geolocations or over legacy protocols by highly privileged accountsazure:signinlogs: sign-in activity; saas:okta: user.authentication.ssoAnomalousLocationThreshold, ProtocolType
AN1504IaaSAPI calls beyond normal scope, such as IAM changes or first use of a serviceAWS:CloudTrail: ConsoleLogin, AssumeRole, ListAccessKeys, CreateUser; gcp:auditServiceInteractionBaseline, RoleSwitchRateThreshold
AN1505SaaSLarge file downloads or new external shares by internal usersm365:unified: FileAccessed, SharingSet; gcp:audit: drive.activityFileDownloadThreshold, SharingPolicyViolationThreshold
AN1506Office SuiteLogins and usage that deviate from the user's normal profilem365:signinlogs: UserLogin; gcp:audit: login.eventBusinessHours, OfficeProductivityToolBaseline

Three lessons follow. First, covering one analytic is not covering the technique: an organisation with strong Entra sign-in analytics but no CloudTrail role-switch baseline covers one of four platforms. Second, the log source names are MITRE's normalised labels, not product table names, so you need a mapping from azure:signinlogs to whatever your SIEM calls that table. Third, every mutable element is a decision with an owner: someone must define the service interaction baseline per role, and keep it current.

What the dataset says about telemetry

Counting log source references across all 1,770 enterprise analytics in v20 shows where detection depends. Use it to sanity-check your collection, then weight it by the techniques that matter to you.

Log source nameReferences in v20 analyticsTypical collection route
WinEventLog:Sysmon872Sysmon with a maintained configuration, forwarded centrally; see the Sysmon Guide
macos:unifiedlog542Unified log streaming or an EDR agent that exports it
auditd:SYSCALL482auditd rules for execve, file and network syscalls; see the Linux Logging and auditd for Defenders Guide
WinEventLog:Security340Advanced audit policy plus Windows Event Forwarding; see the Windows Event Log Guide
NSM:Flow247Network flow records from sensors, firewalls or cloud flow logs
AWS:CloudTrail142Organisation trail to central storage, including data events where analytics need them
m365:unified89Purview unified audit log export to the SIEM
WinEventLog:PowerShell67Script block and module logging (4103, 4104)
esxi:hostd, esxi:vmkernel, esxi:shell143 combinedHypervisor syslog forwarding, often missing entirely
azure:signinlogs36Entra sign-in logs, interactive and non-interactive, plus service principal sign-ins

By data component, DC0032 Process Creation is referenced 866 times, DC0064 Command Execution 339, DC0039 File Creation 254, DC0085 Network Traffic Content 253 and DC0082 Network Connection Creation 247. By platform, the live analytics split as Windows 476, Linux 359, macOS 353, ESXi 107, PRE 97, IaaS 86, Network Devices 86, SaaS 65, Office Suite 57, Identity Provider 46 and Containers 38. Endpoint process telemetry is still the backbone, but cloud, SaaS, identity and hypervisor analytics now exist in numbers that make "endpoint only" programmes visibly incomplete. See the Endpoint Telemetry Guide and the Cloud Logging and Audit Trail Guide.

From analytic to production rule

  1. Pick techniques by threat, not alphabet: start from the techniques your threat model and incident history prioritise, then list their strategies and the analytics for platforms you actually run.
  2. Check telemetry first: for each analytic, confirm every referenced log source and channel is collected, parsed and retained long enough for the time window. Record gaps as collection work, not detection work.
  3. Map names to your schema: maintain a lookup from MITRE log source names and channels to your SIEM tables and fields (for example WinEventLog:Sysmon plus EventCode=10 to your process access table). Normalised schemas reduce this to one mapping per source.
  4. Write the rule in a portable form: express the behaviour in Sigma or your detection-as-code format, tagging it with the technique, DET and AN IDs and the ATT&CK version. See the Sigma Rules Guide and the Detection as Code Guide.
  5. Set the mutable elements explicitly: store each one as a named parameter with its value, rationale and owner, such as a 10-minute TimeWindow or a UserContext exclusion list of approved service accounts.
  6. Test with emulation: run an authorised atomic test or breach and attack simulation for the behaviour, confirm the event arrives and the rule fires, and keep the test as a regression case. See the Purple Teaming Guide.
  7. Shadow, then promote: run the rule in alert-only mode for two to four weeks, review every hit, tune parameters, then promote it with a documented response action. Roll back by disabling the rule version, not by editing it in place.
  8. Record coverage at analytic level: mark the analytic as implemented, tested or tuned, with the date and ATT&CK version, so coverage maps reflect evidence.

Tuning with mutable elements

Across v20 enterprise analytics, 1,669 of 1,770 list at least one mutable element. The most common fields show where local judgement matters most:

  • TimeWindow (581 analytics): the correlation window between events in a sequence. Too short misses slow attacks; too long joins unrelated activity. Set it per analytic from your own event timing, not a global default.
  • UserContext (230): which accounts are in or out of scope, for example administrators, service accounts or break-glass identities. This is where most identity-related false positives are removed, and where attackers hide if exclusions are too broad.
  • ParentProcessName and ParentProcess (85 combined): allow lists of expected parents for a child process, usually build, deployment and management tooling.
  • Entropy thresholds (52 combined): cut-offs for high-entropy domains, command lines or payloads, which vary with your own software estate.
  • Command line patterns and user agents: local strings that define normal administrative use.

Review exclusions as carefully as rules. A UserContext exclusion that covers "all service accounts" removes exactly the non-human identities an intruder most wants to use.

Working with the data programmatically

  • Source: MITRE publishes the STIX 2.1 bundles in the attack-stix-data repository on GitHub, one per domain and version, and points to a TAXII 2.1 server implementation. Pin the version file you used, for example enterprise-attack-20.0.json.
  • Relationships: MITRE's usage document shows strategies linked to techniques by detects relationships and helper functions in both directions that filter out revoked and deprecated objects. Analytics are linked from the strategy's x_mitre_analytic_refs, not by relationships.
  • Filter state: exclude objects where revoked or x_mitre_deprecated is true, and drop references to revoked techniques such as Impair Defenses.
  • Diff each release: compare strategy and analytic x_mitre_version and modified values between releases to find patched logic; v20 alone patched sixteen enterprise analytics.
  • Generate artefacts: from the bundle you can produce a per-technique checklist of log sources, a data collection gap report and a tuning register of mutable elements, rather than maintaining spreadsheets by hand.

Where identity fits

Identity Provider, SaaS, Office Suite and IaaS analytics together number 254 in v20, and many of them watch non-human identities: service principals, OAuth apps, access keys and assumed roles. Two practical points. First, these analytics need identity telemetry that is often collected partially, such as non-interactive and service principal sign-ins or CloudTrail role assumption. Second, baselines for machine identities should be separate from human ones, because a workload's behaviour is narrow and stable, which makes deviation easier to spot. See the Identity Threat Hunting Guide and the MITRE ATT&CK T1078 Valid Accounts Guide. Technique pages in this series apply the model to specific behaviours, for example the T1213 Data from Information Repositories Defensive Guide and the T1555 Credentials from Password Stores Defensive Guide.

Decision guidance

SituationApproachTrade-off
Small team, one SIEM, mostly vendor contentUse MITRE analytics to check vendor rules: for each priority technique, confirm a vendor rule covers each relevant platform analytic and its log sourcesLow effort; you depend on vendor tagging accuracy
Mature detection engineering functionTreat analytics as requirements in a detection-as-code pipeline, with tests and a mutable element registerHigher effort; strongest evidence for coverage claims
Hybrid estate with gaps in cloud and SaaS logsRun the log source aggregation first and fund collection before writing rulesSlower first rules; avoids rules that can never fire
MSSP or multi-tenant SOCKeep a shared rule per analytic with per-tenant mutable element valuesNeeds parameter management tooling

Detection and measurement

  • Telemetry readiness: percentage of analytics for priority techniques whose log sources are all collected and parsed.
  • Analytic implementation rate: analytics implemented and tested, divided by analytics applicable to your platforms, per priority technique.
  • Test pass rate: share of emulation tests that produce an alert within the expected time, re-run after every release and pipeline change.
  • Tuning debt: number of mutable elements left at a default or without an owner.
  • Release lag: days between an ATT&CK release and completing your review of new and patched strategies.
  • Alert precision: true positives divided by alerts per rule over 30 days, to find rules that need parameter work.

Common failure patterns

  • Scoring by data source: data sources are deprecated, so a tool or spreadsheet built on them misses the analytic and data component structure and over-reports coverage.
  • Counting the strategy, not its analytics: marking a technique green because one Windows analytic is live, while its identity and cloud analytics are not.
  • Copying text as a rule: analytics are descriptions; pasting one into a query without the sequence, window and context logic produces noise or nothing.
  • Unowned exclusions: broad UserContext or parent process allow lists added during tuning and never reviewed, which quietly exempt service accounts and automation tooling.
  • Unpinned versions: rules tagged with technique IDs but not the ATT&CK version, so revoked techniques and deprecated strategies linger after a release.
  • Short retention: analytics with long time windows built on sources kept for days, such as some SaaS audit APIs, which cannot support the correlation they need.

Framework mapping

FrameworkRelevant items
NIST SP 800-53 Rev. 5AU-2 Event Logging; AU-12 Audit Record Generation; AU-6 Audit Record Review, Analysis, and Reporting; SI-4 System Monitoring; CA-7 Continuous Monitoring; RA-10 Threat Hunting
NIST CSF 2.0DE.CM-01 networks monitored; DE.CM-09 computing hardware, software and data monitored; DE.AE-02 adverse events analysed; ID.IM-01 improvements identified from evaluations
ISO/IEC 27001:2022 Annex A5.7 Threat intelligence; 8.15 Logging; 8.16 Monitoring activities
CIS Controls v8.1Control 8 Audit Log Management; Control 13 Network Monitoring and Defense
MITRE D3FENDComplements ATT&CK analytics with a defensive technique vocabulary; see the MITRE D3FEND Guide

Practitioner checklist

  • Retire any coverage model or tooling that still scores by ATT&CK data source.
  • Pin the ATT&CK version you map against and record it on every rule.
  • Generate a log source checklist from the analytics for your priority techniques and close collection gaps first.
  • Maintain a mapping from MITRE log source names and channels to your SIEM tables and fields.
  • Implement analytics per platform and record coverage at analytic level, not technique level.
  • Keep a register of mutable element values with rationale and owner, and review exclusions quarterly.
  • Back every implemented analytic with an emulation test run after each release and pipeline change.
  • Review new, patched and deprecated strategies within 30 days of each ATT&CK release.
  • Baseline non-human identities separately in identity, SaaS and cloud analytics.

Standards and references

Related NHI Mgmt Group resources: MITRE ATT&CK Coverage Mapping Guide · Detection Engineering Guide · Sigma Rules Guide · MITRE ATT&CK Guide for Defenders

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 27 October 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.