Coverage scoring rules
The rules behind every color in the coverage app and the coverage workbook. Both apply the same statuses and the same weighting, so a matrix exported from one matches the other.
Do it in the browser Map your ATT&CK coverage
On this page
Statuses
| Status | When a technique gets it |
|---|---|
| Covered | A live detection at robustness 4 or 5, the Summiting the Pyramid levels tied to the technique itself. |
| Partial | Live detections exist, but the strongest is at robustness 1 to 3, tied to a tool or an artifact the attacker controls. A parent technique is also partial when some of its sub-techniques are detected, and covered when its own best detection reaches 4 or 5; a sub-technique never lifts its parent past partial. |
| Ready to build | No live detection, and every log source one of ATT&CK's analytics for the technique reads is collected. Analytics written for platforms you do not run do not count. |
| Some telemetry | No live detection. Some of the log sources the technique's analytics read are collected, but no single analytic has all of them, so the next step is collecting what one analytic still lacks. |
| Blind | No live detection, and none of the log sources the technique's analytics read are collected. |
| Check telemetry | A live detection exists, but none of the log sources the technique needs are marked collected. Either the log source list is out of date or the detection reads a log source nobody collects. Check telemetry wins over covered and partial, and the technique still counts in the detected and weighted shares. |
| Out of scope | Excluded from every count. Techniques whose only platform is PRE start out of scope; anything else can be switched off per technique. |
The coverage numbers
Coverage is four numbers that answer different questions. The two per tactic keep growing; the two per technique cap at the matrix.
| Number | The question it answers | How it is computed | Where it tops out |
|---|---|---|---|
| Detection count | How many live detections name a technique in this tactic? | One per live detection, counted under every tactic it maps to. | Never. A brittle detection adds to it as much as a strong one. |
| Robustness-weighted coverage | How much evasion-resistant work sits in this tactic? | Each live detection counts as its robustness divided by 5, summed per tactic. | Never. Five detections at robustness 1 equal one at robustness 5. |
| Weighted share | What share of the matrix is covered, and how well? | Each technique scores its strongest live detection's robustness divided by 5, averaged across every technique. | 100%, when every technique has a detection at robustness 5. |
| Covered count | How many techniques are covered at all? | A technique counts once its strongest live detection is at robustness 4 or 5. | The technique count of the pinned ATT&CK release. |
Scope
- Pick the platforms you run, and techniques that exist only on other platforms go out of scope. Leave every platform unticked to keep them all in scope. Any technique can be switched in or out by hand, and a hand setting wins.
- Techniques whose only platform is PRE, the reconnaissance and resource development a defender cannot log, start out of scope.
Log sources
- Log sources are listed by the names ATT&CK's own analytics use, such as WinEventLog:Sysmon, auditd:SYSCALL, AWS:CloudTrail, and m365:unified, grouped by the platform they serve. Ticking one marks every kind of event it produces as collected.
- Behind the names sit ATT&CK's data components, the kinds of event a detection strategy reads. Each can be ticked or unticked on its own, and a hand setting overrides what the named log sources imply.
- A log source's rank is the number of techniques it would make ready to build by completing one of their analytics. A kind of event ranks by the analytics it is the last missing piece of.
Weighting and counts
- A technique's weighted coverage is its strongest live detection's robustness divided by 5. Extra detections at a lower level add nothing, which matches how the detect score treats a candidate the live set already covers.
- Detected share counts any live detection; weighted share uses the robustness weighting. When the weighted share is much lower, most detections in that tactic match tools and artifacts an attacker can change.
- Weighted share is not the program metric robustness_weighted_coverage. That metric adds robustness divided by 5 for every live detection in a tactic, so it keeps growing as detections are added. Weighted share caps each technique at its strongest detection, so it reads as a percentage of the matrix.
- Only detections with status live count. Live means deploy or tune, the two stages the kit counts as live, so a detection running with alerting off counts. Building covers hypothesis through validate. Building and retired detections stay in the list so the pipeline and the workbook can see them.
- The suggested telemetry readiness for a technique is the lowest readiness among the log sources of its best complete analytic, scored on the Log sources tab. It is 1 when no analytic is complete, and 0 when none of its log sources is collected. When ATT&CK lists no log source, the app suggests none, because only your inventory can say.
Working with it
- Import detection rules from where they already live. The app reads a folder of Sigma rules, taking technique ids from their attack tags. It also reads Splunk savedsearches.conf or its REST export, Microsoft Sentinel rule exports in YAML or JSON, and Elastic rule exports in ndjson. Rules do not say how hard they are to evade, so each arrives unscored and counts as partial until someone scores it. An unscored detection counts as robustness 0 in the weighted share and the Navigator score.
- Robustness is scored with four questions from MITRE CTID's Summiting the Pyramid, one per level: changeable values, attacker tools, environment tools, and variant-specific artifacts. Does the detection match a value the attacker changes in seconds? A tool the attacker brings? A tool already in the environment? Something only some ways of doing the technique produce, so another way gets past? The number of the earliest question answered yes is the level, 1 to 4; four noes make it 5.
- ATT&CK Navigator layers go both ways. The export colors every technique by status and scores it by its strongest live detection's robustness. An imported layer shows as an overlay, so a threat-intelligence layer can sit on top of the coverage.
- Threat groups come from ATT&CK's own records of which groups used which techniques. Choosing the groups that matter to you marks their techniques and puts them first in what to build next.
- A snapshot stores the date, the totals, the tactic shares, and every technique's status. Comparing with an older snapshot shows the change per tactic and, within one ATT&CK version, which techniques changed status.
Sharing
- Merging a teammate's backup adds the detections missing from yours, adds the logs they ticked, and asks which version to keep when the same detection id differs.
- A read-only view link carries every technique's status packed into the part of the address after #, which browsers do not send to the web server. It carries no detection names, log sources, or notes. It still shows where you are blind, so share it the way you would share the report.
- A link only opens against the ATT&CK version it was made with. After an update it says so and asks for a new link, since the technique list may have changed underneath it.
ATT&CK versions
- The site pins the ATT&CK release it was built with. A daily job checks MITRE's index and rebuilds the site when a new release lands.
- Detections are saved with the release they were entered against. After an update, the app lists every saved reference that MITRE revoked (with the replacement MITRE names), deprecated, or moved to another tactic. Nothing changes until you accept it.
Privacy
- Everything you enter is stored in your browser's own database on your machine. The page sends nothing anywhere.
- Every change in the app reports what moved and can be undone from its message or with Ctrl or Cmd+Z, one step back, within the matrix it was made in.
- The one outbound request is the check of MITRE's public release index, which stays off until you turn it on, runs at most once a day, and carries nothing about your data.
- Clearing your browser's site data deletes the workspace. Back up to a file, which is also how a teammate gets a copy.
Open the coverage app, or download the workbook to do the same in Excel.