# KEV states what has been exploited and EPSS estimates what will be; on listing day most KEV entries scored below 0.1

COMPARISON, read 10 Nov 2026

**The short answer.** The CISA KEV catalog lists CVEs with confirmed exploitation; EPSS estimates the chance a CVE will be exploited in the next 30 days. FIRST advises treating any KEV entry as exploited regardless of its EPSS score. In the KEV Lag Ledger, the median EPSS on listing day was 0.0066, and 75.8 percent of 260 scored listings were below 0.1.

- **Maintained by:** EPSS: FIRST EPSS SIG, scores generated by Empirical Security. KEV: CISA
- **Scale:** EPSS: probability 0 to 1. KEV: listed or not listed
- **Current version:** EPSS model v2026.06.15. KEV catalog 2026.11.10, 1,779 entries
- **Updated:** EPSS: daily. KEV: on each addition
- **Answers:** EPSS: the chance of exploitation in the next 30 days. KEV: whether exploitation is confirmed

## Prediction and record

KEV is a record: CISA lists a CVE when it holds reliable evidence that the exploit is being actively used. EPSS is a prediction: a probability, published daily for every CVE, that exploitation will be seen in the next 30 days. A record covers the past and a prediction the near future, and the two can disagree without either being wrong.

FIRST says so in its Using EPSS guidance: a confirmed exploitation report and a low EPSS score are not automatically contradictory, a low score means the signal profile resembled low-risk vulnerabilities right up to the event, and confirmed and recent exploitation evidence should take precedence over the score when it exists.

| Property | EPSS | KEV |
| --- | --- | --- |
| Nature | Prediction, calibrated probability | Record of confirmed exploitation |
| Scope | Every published CVE | 1,779 CVEs on 2026-11-10 |
| Time | Next 30 days | Already happened |
| Refresh | Daily | On each addition |
| Failure mode | Low score on a CVE that is then attacked | Silent until CISA lists |

## The listing-day record

The KEV Lag Ledger asks what EPSS said on the day CISA listed each CVE, taken from the FIRST API for that date. Of 318 additions, 260 had a score. The median was 0.0066 (0.006555). 197 (75.8 percent) were below 0.1, 116 (44.6 percent) were at 0.01 or above, 63 (24.2 percent) at 0.1 or above and 34 (13.1 percent) at 0.5 or above. 75 (28.8 percent) were in the 90th percentile or higher.

The 58 additions with no EPSS value that day had a median gap of 0 days from NVD publication, so they were CVEs listed on or near the day they appeared. For them EPSS had not yet scored the CVE. A threshold-only policy has no answer for that group, while a KEV check does.

The rank view tells the same story. 75 of the 260 scored additions (28.8 percent) sat in the 90th percentile or higher on listing day, and 55 (21.2 percent) in the 95th or higher. Three in ten of the confirmed set were therefore in the top tenth of the distribution, and seven in ten were not.

## Newest and oldest additions

The ledger separates two groups. Of the 86 scored additions listed within 7 days of publication, 81 (94.2 percent) had EPSS below 0.1. The FIRST page on limits explains why: for a newly disclosed vulnerability the public signal is thin, exploit code may not exist, and EPSS still produces a score but from a sparse record.

The 55 additions listed more than a year after publication look different: median EPSS on listing day was 0.347, and 23 of 55 were still below 0.1. Old CVEs accumulate signal, so EPSS tends to catch up, but it is not complete even then.

The 29 additions marked with known ransomware use give a smaller sample: 25 of them had a listing-day score, with a median of 0.0114. That is higher than the 0.0066 median of all scored entries and still far below 0.1, so EPSS did not single out the ransomware-linked entries on the day they were listed.

## How EPSS and KEV feed each other

The two are linked. FIRST names VulnCheck KEV, the Shadow Server Foundation and LevelBlue among the data partners for exploitation activity. FIRST also states that exploitation activity trains the model and does not drive the daily score, so a CISA listing does not push tomorrow's EPSS up. The Using EPSS page adds that a vulnerability added to KEV years ago reverts toward the probability its current features predict.

FIRST's rule of thumb is explicit: when a vulnerability appears on CISA KEV, treat it as actively exploited and prioritize accordingly, regardless of EPSS score. EPSS remains useful for everything KEV does not cover, which FIRST describes as the large majority of the vulnerability population at any given time.

## A combined rule

Use KEV as the override and EPSS as the ranking for the rest. For CVEs not in KEV, choose an EPSS threshold from the distribution: FIRST says the base rate of exploitation is roughly 2 to 3 percent and that the top half of published vulnerabilities sits above about 0.7 percent, so a threshold set by intuition, such as 50 percent, sits far above the median score.

Review the queue against KEV each week. The ledger shows 17 to 43 additions in each full month, and the CVEs that enter KEV tomorrow are often ones EPSS ranked low today. A weekly check of which KEV entries had been in your EPSS-ranked queue tells you how the threshold performed on your assets.

## Reporting both

Show EPSS as a probability with its percentile, for example 15.3% (92nd), as FIRST recommends, and show KEV as a dated yes or no. Do not display a combined number.

The ledger is the baseline for the report: it states that on listing day a 0.1 threshold would have flagged one in four of the confirmed entries. A program can adopt a lower threshold to raise coverage, at the cost of effort, and should record that choice and the reason.

If the two disagree on one CVE, write down which one decided. A KEV entry with a low EPSS is decided by KEV. An unlisted CVE with a high EPSS is decided by EPSS and reviewed by a person. The record of those disagreements, kept for a quarter, is the most useful evidence for adjusting a threshold.

## Verdict by situation

- **Federal agencies and contractors:** Rank by KEV first because BOD 26-04 requires it, and use EPSS only to order the CVEs that are not yet in the catalog.
- **Large enterprises with ten or more scanners:** Rank by KEV first as the override and EPSS second as the queue for everything else, and review weekly which new KEV entries EPSS had ranked low.
- **Managed security providers:** Rank by KEV first because it is the one list every client accepts without argument, then by EPSS for the open backlog in each client.
- **Mid-size teams replacing Kenna or a spreadsheet:** Rank by KEV first because it is free and unambiguous, then by EPSS to replace the exploit-likelihood part of the old risk score.
- **Application security and infrastructure in one queue:** Rank by KEV first for infrastructure and for any application component with a CVE, and use EPSS for open-source dependencies, which carry CVE identifiers.

## Other comparisons

- [EPSS vs CVSS](/compare/epss-vs-cvss/)
- [CVSS vs KEV](/compare/cvss-vs-kev/)
- [SSVC vs CVSS](/compare/ssvc-vs-cvss/)
- [SSVC vs EPSS](/compare/ssvc-vs-epss/)
- [CVSS v3.1 vs v4.0](/compare/cvss-v3-1-vs-v4-0/)
- [KEV vs vendor intelligence](/compare/kev-vs-vendor-intelligence/)
- [CVSS vs EPSS vs KEV](/compare/cvss-vs-epss-vs-kev/)

## Sources

- FIRST, EPSS home (data partners, maintenance): https://www.first.org/epss/ (read 10 Nov 2026)
- FIRST, Using EPSS (KEV guidance, thresholds): https://www.first.org/epss/using-epss (read 10 Nov 2026)
- FIRST, How EPSS works (limitations): https://www.first.org/epss/how-it-works (read 10 Nov 2026)
- CISA, KEV catalog JSON feed: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 10 Nov 2026)
- CISA, BOD 22-01 (superseded), archived copy: https://web.archive.org/web/2025/https://www.cisa.gov/news-events/directives/bod-22-01-reducing-significant-risk-known-exploited-vulnerabilities (read 10 Nov 2026)

Last reviewed 10 Nov 2026. Canonical: https://epssvscvss.com/compare/epss-vs-kev/
