Skip to content
EPSS vs CVSS

COMPARISON, read 10 Nov 2026

Rank by KEV first, EPSS second and CVSS third; each answers a different question and the KEV Lag Ledger shows why one alone fails

All comparisonsThe ledger

Maintained by
CVSS: FIRST CVSS-SIG. EPSS: FIRST EPSS SIG. KEV: CISA
Scale
CVSS: 0.0 to 10.0. EPSS: 0 to 1. KEV: listed or not
Current version
CVSS v3.1 and v4.0. EPSS v2026.06.15. KEV catalog 2026.11.10
Updated
CVSS: on scoring. EPSS: daily. KEV: on addition
Answers
Severity, likelihood and confirmation, in that order of availability

Three questions

CVSS answers how bad the flaw is, EPSS how likely exploitation is in the next 30 days, and KEV whether exploitation has been confirmed. Each has a published source: the FIRST specifications, the FIRST EPSS pages and daily CSV, and the CISA JSON feed. They are available at different times: CVSS near publication, EPSS daily from the first days, KEV after CISA holds evidence.

A ranking that uses one of the three answers one question. The ledger shows what is lost in each case.

PropertyCVSSEPSSKEV
QuestionHow severeHow likely in 30 daysIs it confirmed exploited
Output0.0 to 10.0Probability and percentileYes or no, with dateAdded
AvailableNear publicationDailyAfter confirmation
ReproducibleFormula or lookupDocumented, pipeline unpublishedPublic feed
Ledger signal10.4% of 318 under 7.075.8% of 260 under 0.1Median 10.5 days after NVD publication

What the ledger shows for each

The KEV Lag Ledger holds 318 CISA KEV additions from 2025-11-10 to 2026-11-09. On CVSS: median 8.8, 89.6 percent at 7.0 or above, 45.9 percent at 9.0 or above, 10.4 percent below 7.0. On EPSS: 260 with a listing-day score, median 0.0066, 75.8 percent below 0.1, 24.2 percent at 0.1 or above. On KEV: median 10.5 days from NVD publication to listing, 45.3 percent within 7 days, 60.4 percent within 30 and 17.3 percent more than 365.

Cross CVSS and EPSS for the 260: 63 had CVSS 7.0 or above and EPSS 0.1 or above, 173 had CVSS 7.0 or above and EPSS below 0.1, 24 had neither, and none had EPSS 0.1 or above with CVSS below 7.0. At a CVSS cutoff of 9.0 the cells are 35 (both), 88 (CVSS only), 28 (EPSS only) and 109 (neither). At the 7.0 cutoff EPSS found no exploited CVE that CVSS missed; at 9.0 it found 28.

The month counts add scale. Additions per month in the ledger window ran from 17 in 2026-01 to 43 in 2026-09 across full months, with 29 of 318 marked as used by ransomware campaigns. These are public figures from the CISA feed and can be recomputed from the ledger file.

The order and the reasons

First, KEV. It states a fact, it is public, and BOD 26-04 requires it. A KEV-listed CVE on an asset goes to the front of the queue, with the deadline set by the exposure and impact answers in the CISA decision tree. FIRST gives the same instruction for EPSS: treat a KEV entry as actively exploited regardless of its score.

Second, EPSS. It ranks the unlisted population by likelihood, daily, and 58 of the 318 additions had no score on their listing day, so a rule must say what happens when EPSS is missing. Third, CVSS. It sets the floor by asset tier and supplies the severity statement an auditor expects.

The order reflects evidence strength about the present. It does not mean CVSS is the least useful. In the ledger, CVSS at 7.0 flagged 89.6 percent of listings, more than EPSS at 0.1, which flagged 24.2 percent of the scored ones. It says that a flag from KEV needs no further argument and a flag from CVSS needs a second signal.

Failure modes

Using KEV alone is reactive: it lists after exploitation is confirmed, 10.5 days after NVD publication at the median and a year or more after for 17.3 percent. Using EPSS alone misses the many listings that scored low on the day, 197 of 260. Using CVSS alone floods the queue with severity and still leaves 33 listings below 7.0.

Blending the three into one number is the fourth failure. FIRST calls multiplying EPSS by an ordinal score laundering. Keep the three signals in separate columns and let the rule, not a formula, decide the order.

A fifth point concerns time. Each signal ages differently. A CVSS score changes only when someone rescores. An EPSS score changes daily. A KEV listing persists, but FIRST notes that its value as a predictor of continued exploitation fades without other evidence of ongoing activity. Policies should state how often each signal is re-read, and the date on each report should be the date of the read.

A rule set you can audit

Write the rule as four lines. If the CVE is in KEV, apply the KEV timeline. Else if EPSS is at or above the documented threshold, apply the high-likelihood timeline. Else if CVSS is at or above the documented cutoff for the asset tier, apply the severity timeline. Else apply the standard cycle.

Store, for each decision, the three values, the versions, the sources and the read date: KEV catalogVersion, the EPSS model version from the CSV header, and the CVSS version and vector.

Keep the rule short enough to apply in a meeting. If three people read the rule and reach different orders for the same ten CVEs, it needs rewriting. The ledger rows are a convenient test set: take ten, apply the rule, and compare.

What this site does not cover

The ledger measures public signals on listing day. It does not measure presence on your assets, reachability, compensating controls or business impact, which are the inputs that SSVC and platform vendors add. Methods that test exploitability in your own environment are out of scope here.

Re-run the ledger method on your own data each quarter. Its sources are the CISA KEV JSON, the NVD API 2.0 and the FIRST EPSS API.

Verdict by situation

  • Federal agencies and contractors: Rank by KEV status first, as BOD 26-04 requires, then by the directive's exposure and impact answers, and use EPSS and CVSS only for what KEV does not list.
  • Large enterprises with ten or more scanners: Rank by KEV first, EPSS second and CVSS as the tier floor, with the three values stored in separate columns so each decision can be defended.
  • Managed security providers: Rank by KEV first and EPSS second across every client, because both are public and comparable, and show CVSS in each client report.
  • Mid-size teams replacing Kenna or a spreadsheet: Rank by KEV first and EPSS second, which replaces the exploit-likelihood part of the old risk score with two free public inputs, and keep CVSS for severity.
  • Application security and infrastructure in one queue: Rank by KEV first and EPSS second for CVE-bearing infrastructure and dependencies, and by CVSS first for code-level application findings that have no CVE.

Other comparisons

Sources

  1. CISA, KEV catalog JSON feed, read 10 Nov 2026
  2. FIRST, CVSS v3.1 specification, read 10 Nov 2026
  3. FIRST, CVSS v4.0 specification, read 10 Nov 2026
  4. FIRST, EPSS home, read 10 Nov 2026
  5. FIRST, Using EPSS (KEV rule, score laundering), read 10 Nov 2026
  6. CISA, BOD 26-04, archived copy, read 10 Nov 2026
  7. CERT/CC, CISA BOD 26-04 Response Model, read 10 Nov 2026

Last reviewed 10 Nov 2026.