Most SOCs cannot say which attacker behaviours they are blind to, because nobody has mapped the logs they collect to the data those detections need. MITRE ATT&CK v20 makes that mapping practical: each data component (such as Process Creation or User Account Authentication) lists the log sources and channels that supply it, for example WinEventLog:Security event 4688 or kubernetes:audit create events. This template gives detection engineers and SOC leads a ready-to-use matrix and an operating procedure for keeping it current: when to update it, who owns each source, how to record coverage, what to do about gaps and source outages, and how to report the result. It complements the MITRE ATT&CK Coverage Mapping Guide, which covers technique-level coverage, by working one layer down at the log source.
How to use this template
- Copy the sections below into your SOC runbook and replace every [placeholder] with your own values. Keep the section numbers so that review records can refer to them.
- Build the matrix in a spreadsheet or database using the columns in section 3. One row per data component and log source pair, not per technique.
- Use ATT&CK v20 as the reference. The data component list and each component's log sources come from the ATT&CK data set (the
x_mitre_log_sourcesfield in the STIX data); v20 has 106 active data components across enterprise platforms. Record the ATT&CK version in the matrix header so that later updates can be compared. - Start small: the 20 to 30 data components that feed most of your detections, then extend.
- Treat sections 4 to 6 as the gap and outage procedure: they apply when a needed source is missing, degraded or lost, which is the log source equivalent of an incident.
- Review quarterly and after every significant incident or platform change.
1. Scope and triggers
1.1 Scope. This procedure covers every log source that [Organisation] collects, or should collect, for security monitoring across [on-premises, cloud, SaaS, identity, endpoint, network and OT] environments, and the mapping of those sources to ATT&CK data components.
1.2 Out of scope. Application performance and debug logs not used for security, unless a detection depends on them.
1.3 Triggers for an update. The matrix owner updates the matrix when any of the following occurs:
- A new platform, SaaS application, cloud account or cluster enters production.
- A log source is added, removed, re-licensed or changes format or retention.
- A new ATT&CK version is released (review within [60] days).
- A detection is created, retired or found to be broken.
- An incident, red team or purple team exercise reveals missing or late data.
- A source outage of more than [4 hours] on a tier-one source.
2. Roles and contacts
| Role | Responsibility | Contact |
|---|---|---|
| Matrix owner | Maintains the matrix, runs quarterly reviews, reports coverage | [Detection engineering lead] |
| Source owner (one per source) | Keeps the source enabled, configured, licensed and forwarding; notifies changes | [Platform or application owner] |
| Pipeline owner | Collectors, parsers, SIEM ingestion and health monitoring | [SIEM engineering] |
| Collector credential owner | API tokens, service principals and service accounts used to pull logs | [Identity or platform team] |
| SOC manager | Accepts residual gaps, prioritises gap closure | [Name] |
| Risk owner | Signs off accepted gaps on tier-one assets | [CISO or delegate] |
2.1 Every row in the matrix names a source owner. A source without an owner is recorded as a gap.
2.2 Collector identities are non-human identities with tenant-wide read access. They are inventoried, scoped to read-only audit access, and rotated on [schedule]. See the SaaS API Token Security Guide.
3. Detection and triage
3.1 Matrix columns. Record each data component and log source pair with these fields.
| Column | Content | Example |
|---|---|---|
| Data component | ATT&CK v20 name and ID | Process Creation |
| ATT&CK log source and channel | As listed by ATT&CK | WinEventLog:Security, WinEventLog:Sysmon EventCode=1 |
| Our source | Product, feature and table in your SIEM | [EDR process events, SIEM table name] |
| Platform scope | Where the source is deployed | [Windows servers: 96 percent of fleet] |
| Status | Collected, Partial, Planned, Not collected, Not applicable | Partial |
| Quality | Fields present, parsing, time sync | [Command line present: yes] |
| Latency | Typical event-to-SIEM delay | [2 minutes] |
| Retention | Hot and total | [90 days hot, 12 months total] |
| Licence dependency | Tier or add-on required | [Premium audit tier] |
| Detections using it | Rule IDs | [DR-0142, DR-0187] |
| Source owner | Named person or team | [Endpoint team] |
| Last validated | Date of last test event seen | [2026-10-01] |
3.2 Example rows. These show the pattern; ATT&CK v20 lists the log sources in the middle column for each component.
| Data component | ATT&CK v20 log sources (selection) | Typical gap |
|---|---|---|
| Process Creation | WinEventLog:Security, WinEventLog:Sysmon, auditd:SYSCALL, macos:unifiedlog | Linux and macOS not covered; command line not captured |
| User Account Authentication | WinEventLog:Security, AWS:CloudTrail, Okta:SystemLog, GCPAuditLogs:login.googleapis.com, auditd:USER_LOGIN | SaaS identity providers not ingested |
| Logon Session Creation | WinEventLog:Security, azure:signinlogs, CloudTrail:Signin, Okta:SystemLog | Service principal and workload sign-ins excluded |
| Cloud Service Modification | AWS:CloudTrail, azure:activity, gcp:audit, m365:unified, saas:github, saas:slack | Only one cloud collected; SaaS admin logs missing |
| Network Traffic Flow | AWS:VPCFlowLogs, NetFlow:Flow, NSM:Flow, NSM:Firewall | Flow logs off in some VPCs |
| Pod Creation | kubernetes:audit create | Managed cluster audit export not enabled. See the Kubernetes Audit Log Detection Guide |
| Command Execution | WinEventLog:PowerShell, auditd:EXECVE, EDR:cli, AWS:CloudTrail | PowerShell script block logging off |
3.3 Triage of a new requirement. When a detection engineer needs a data component, the matrix owner checks the row: if Collected, the rule proceeds; if Partial or Not collected, sections 4 and 5 apply.
3.4 Status rules. Mark a row Collected only when a test event has been seen in the SIEM within the last [90] days and required fields parse correctly. Coverage on paper does not count.
4. Containment
This section covers how to contain the risk of a gap until it is closed.
4.1 Classify the gap as Missing (source not collected), Partial (some platforms or fields), Degraded (latency or loss above threshold) or Licence-gated (events require a higher tier).
4.2 Compensating sources. Use the matrix to find another log source for the same data component. For example, if endpoint process telemetry is missing on a server group, Windows Security event 4688 with command line auditing may cover Process Creation temporarily; if a SaaS application's admin log is unavailable, identity provider sign-in logs cover part of User Account Authentication.
4.3 Preventive controls. Where no compensating source exists, record the preventive controls that reduce exposure (for example admission control blocking privileged pods while Kubernetes audit export is being enabled).
4.4 Accept or escalate. Gaps on tier-one assets with no compensating source are escalated to the risk owner within [5] working days.
5. Eradication
This section covers closing gaps permanently.
- Prioritise gaps by the number of high-priority techniques that depend on the data component and by asset tier. Use threat intelligence on actors relevant to [Organisation] to weight techniques.
- Raise a work item for the source owner with the required configuration (audit policy, diagnostic setting, licence, collector).
- Onboard in audit mode: forward to a staging index first, confirm volume, parsing and cost, then promote to production.
- Validate with a test event and record the date in the Last validated column.
- Update status and link the detections that now use the source.
5.1 Licence-gated gaps go to the SOC manager with a cost and benefit note listing the detections unblocked. See the SaaS Audit Log Retention and Licensing Guide.
6. Recovery
This section covers a source that was collected and stopped.
6.1 Detect the outage. The pipeline owner monitors every tier-one source for silence beyond [its normal interval plus documented lag], parsing failure rates above [2 percent], and volume drops over [50 percent] against a seven-day baseline.
6.2 Treat unexplained loss as a possible attack. Disabling logging maps to ATT&CK T1685 Disable or Modify Tools. The SOC checks who last changed the source's configuration before assuming a technical fault.
6.3 Restore and backfill. Fix the cause (expired collector credential, API throttling, agent removal, licence lapse), then re-pull the missed window from the platform's native retention where possible.
6.4 Record the gap window in the matrix and in the SIEM so that investigations covering that period know data is incomplete.
7. Communication and notification
7.1 Monthly coverage report to the SOC manager: percentage of priority data components Collected, list of Partial and Not collected, outages and their duration.
7.2 Quarterly report to the risk owner: coverage trend, accepted gaps with expiry dates, licence decisions needed.
7.3 Change notification. Source owners notify the matrix owner [10] working days before any planned change to a source's configuration, licence, format or retention.
7.4 Audit evidence. The matrix and its history serve as evidence for logging controls such as NIST SP 800-53 AU-2 and AU-12, CIS Controls v8.1 Control 8, and ISO/IEC 27001:2022 Annex A 8.15. See the CIS Control 8 Audit Log Management Guide.
8. Post-incident review
8.1 After every incident and purple team exercise, the matrix owner records which data components were needed, whether they were available for the whole period, and how long they took to reach the SIEM.
8.2 Any component that was missing, late or retained too briefly becomes a gap item under section 5, with the incident reference.
8.3 Detections that failed because of data quality (missing fields, parsing errors) are recorded against the row's Quality column.
8.4 The review is complete when the matrix is updated and the owner has signed off.
Implementation guidance
- Generate the skeleton from the data: extract data components and their log sources from the ATT&CK v20 STIX bundle rather than typing them. Each data component object carries
x_mitre_log_sourcesentries with anameandchannel, and ATT&CK v20 analytics link log source references to detection strategies. - Map names to your reality: ATT&CK log source names are generic (
m365:unified,saas:okta). Add your own column with the SIEM table or index name. - Do not chase 100 percent: ATT&CK lists log sources for many platforms you may not run. Mark them Not applicable rather than leaving them blank.
- Combine with technique coverage: link each row to the detections that use it, then roll up to technique coverage in an ATT&CK Navigator layer. See the ATT&CK Detection Strategies and Analytics Guide.
- Include identity sources early: identity provider, cloud control plane and SaaS audit logs are where non-human identity abuse shows up, and they are often the least covered.
- A starter set of priority components: Process Creation, Command Execution, User Account Authentication, Logon Session Creation, User Account Modification, Cloud Service Modification, Network Traffic Flow, File Creation, Windows Registry Key Modification and Pod Creation cover a large share of common detections across endpoint, identity, cloud and container environments. Begin with these rows, validate them, then expand by platform.
- Record the channel, not just the source: collecting
WinEventLog:Securityis not the same as collecting the specific event codes a component needs. Audit policy settings decide which events exist, so the Quality column should name the audit subcategories or policy settings that must be on. - Metrics to track: percentage of priority data components Collected; number of tier-one sources without an owner; mean time to detect a source outage; rows validated in the last 90 days; detections blocked by missing data.
- Small organisations: a single spreadsheet with 20 rows and a quarterly review is enough. Large organisations should store the matrix in the detection-as-code repository and update it through pull requests.
Related NHI Mgmt Group resources: MITRE ATT&CK Coverage Mapping Guide · Detection Engineering Guide · Log Management and Retention Guide · Detection Engineering Maturity Model