Cybersecurity KRIs: Measuring Risk, Not Activity

Performance metrics tell you the team is busy. Risk indicators tell you whether the organisation is getting safer. Here are 24 of each, paired, with the thresholds that turn a number into a decision.

12security domains
24KPIs — is the work being done?
24KRIs — is exposure building up?
3threshold bands: green, amber, red
2 wksCISA deadline for newer known-exploited CVEs

A security dashboard full of green KPIs can sit on top of a growing pile of risk. “98% of critical vulnerabilities closed within SLA” is a good KPI; it says nothing about the one internet-facing server with a vulnerability attackers are exploiting today. That second fact is a key risk indicator (KRI). KPIs measure the performance of controls. KRIs measure the exposure the controls are supposed to reduce, and they are tied to risk appetite: each one has a threshold above which someone must act.

This page extends Cybersecurity KPIs, which covers performance metrics and the MTTD/MTTR clocks.

1 · KPI vs KRI in One Table

KPIKRI
QuestionAre we doing the work well?How exposed are we right now?
Typical formRate, coverage %, time to completeCount of open exposures, age of the worst ones
Time senseMostly lagging — what got doneLeading — what could go wrong next
Anchored toTeam targets and SLAsBoard-approved risk appetite
When it turns redFix the processEscalate the risk to its owner
ExampleEDR coverage across managed endpoints: 97%Endpoints with no active protection: 41

▸ The pair often measures the same thing from opposite ends. 97% coverage and 41 unprotected endpoints can be the same fact — but only the KRI tells you the 41 includes three domain controllers.

2 · 24 Pairs Across 12 Domains

Two KPIs and two KRIs per domain. Read across a row: the KPI is the effort, the KRI is what that effort is supposed to drive down.

DomainKPIs (performance)KRIs (exposure)
IdentityPhishing-resistant MFA coverage · privileged-access review completionPrivileged accounts without strong MFA · stale or orphaned accounts still active
Attack surfaceInternet-facing asset inventory coverage · exposed-service remediation timeUnknown internet-facing assets found · critical services exposed to the internet
VulnerabilitiesKnown-exploited vulnerability remediation time · critical vulnerability closure rateKnown-exploited vulnerabilities open · ageing critical backlog
ConfigurationCritical configuration compliance rate · high-risk misconfiguration closure rateHigh-risk misconfigurations outstanding · unauthorised configuration drift
EndpointEDR coverage · endpoint isolation response timeEndpoints without active protection · malware containment delays
Email & socialReported-phish triage time · malicious email blocking rateConfirmed credential-phishing incidents · business email compromise attempts
NetworkSegmentation control coverage · time to close unapproved inbound exposureUnsegmented paths to critical systems · unauthorised open ports or services
Application & APISecure code review completion · API security test coverageCritical application flaws in production · unauthenticated or weakly protected APIs
CloudCloud finding remediation time · least-privilege access review completionPublicly exposed cloud resources · excessive cloud privileges outstanding
DataSensitive-data discovery coverage · data-loss alert resolution timeSensitive data exposed or misrouted · unencrypted critical data stores
Detection & responseMean time to detect · mean time to containHigh-severity alerts past SLA · confirmed incidents growing in scope
ResilienceBackup restore-test success rate · recovery exercise completionRestore failures during testing · recovery objectives (RTO/RPO) missed

3 · Thresholds Turn Numbers into Decisions

A KRI without a threshold is just another number. Each one needs three bands, an owner, and a named action when it crosses into amber or red. The values below are illustrative starting points for a mid-sized organisation; your own should come from your risk appetite statement.

KRIGreenAmberRed — escalate
Known-exploited vulns open on internet-facing systems01–2, under 7 days oldAny older than 14 days
Privileged accounts without phishing-resistant MFA01–3 with an approved exceptionAny without an exception
Unknown internet-facing assets found this month0–23–10More than 10, or any hosting data
Publicly exposed cloud storage0—Any not on the approved public list
High-severity alerts past SLAUnder 2%2–5%Over 5%, or any on a crown-jewel system
Restore failures in last quarter’s tests01 non-critical systemAny critical system

▸ Notice how many reds are “any”. For exposures that attackers actively hunt, the right threshold is often zero, with a documented exception process — not a percentage that hides the one that matters.

4 · The KEV Clocks

Most organisations cannot patch everything, so the vulnerability KRIs should count what is actually being exploited, not everything a scanner reports. The public reference is CISA’s Known Exploited Vulnerabilities (KEV) catalog. Its operating directive for US federal agencies, BOD 22-01, sets two remediation clocks that work well as a benchmark anywhere:

VulnerabilityBOD 22-01 deadlineUse it as
KEV entry, CVE assigned 2021 or later2 weeksThe red line for internet-facing systems
KEV entry, older CVE6 monthsAn outer limit — most teams should be far faster
Not in KEV, high exploit probabilityNo federal clockRank with FIRST’s EPSS score plus asset criticality

CISA can set a shorter deadline for an individual entry. EPSS (Exploit Prediction Scoring System, from FIRST) estimates the probability a vulnerability will be exploited in the next 30 days — a useful second sort after KEV.

5 · What Goes to the Board

The board does not need 48 numbers. It needs to know whether exposure is inside appetite, which KRIs are red, who owns them and by when they will be green.

  1. Pick 6–8 KRIs tied to your top risks — typically KEV exposure, privileged MFA gaps, unknown external assets, public cloud exposure, alerts past SLA and restore failures.
  2. Show the trend over the last four quarters, not just today’s value.
  3. Every red gets an owner and a date — or a formally accepted risk signed by that owner.
  4. Pair with two KPIs — MTTD and MTTR — so the board also sees whether the team is getting faster.

Bottom line

  • KPIs measure effort, KRIs measure exposure — you need both, and boards need mostly the second.
  • Pair them — every KPI should drive a KRI down; if it doesn’t, question the KPI.
  • Thresholds or it isn’t a KRI — green, amber, red, an owner and an action.
  • Zero is a valid threshold for what attackers actively hunt.
  • Sort by exploitation — KEV first, then EPSS and asset criticality.

Sources & method. Domains and metric names reflect common security-metrics practice. KEV deadlines from CISA Binding Operational Directive 22-01 (November 2021); EPSS from FIRST. Threshold values are illustrative, not a standard, and should be set from your own risk appetite. Original analysis; no third-party graphics or text reproduced.

← Back to Cyber security