# CVSS measures severity, so ranking by it alone puts exploited CVEs in the wrong place

LEDGER FINDING, read 10 Nov 2026

**The short answer.** CVSS describes how severe a flaw is, not whether anyone is exploiting it. In the KEV Lag Ledger, 10.4% of 318 CISA KEV additions scored below 7, and the median EPSS on listing day was 0.0066. Each signal misses a share of exploited CVEs, which is why the signals must be combined.

## What CVSS was built to say

FIRST's specification describes the base score as reflecting the severity of a vulnerability according to its intrinsic characteristics, and places factors such as monetary losses due to a breach outside the scope of CVSS. A score of 9.8 describes the flaw. It does not describe your asset, your controls or the attackers' interest.

The v4.0 document asks consumers to enrich Base with Threat and Environmental metrics. Most published scores are base only, so the number you receive is the least informed version.

## What the ledger shows

The KEV Lag Ledger covers every KEV addition from 2025-11-10 to 2026-11-09: 318 CVEs, each with an NVD base score. The headline figures are below, each computed from the rows in data/kev/ledger.json.

Read them together. Almost half of the exploited CVEs would top a CVSS ranking. About one in ten would sit below the High band. The probability signal, on the day of listing, was low for most. No single column reliably flags the exploited set.

| Measure | Value | Count |
| --- | --- | --- |
| KEV additions in the window | 318 | 318 |
| CVSS base score below 7 | 10.4% | 33 of 318 |
| CVSS base score 9.0 or above | 45.9% | 146 of 318 |
| CVSS base score of 10 | 9.4% | 30 of 318 |
| EPSS below 0.10 on listing day | 75.8% | 197 of 260 |
| Median EPSS on listing day | 0.0066 | 260 with a score |
| Median days, NVD publication to listing | 10.5 | 318 |
| Listed within 7 days of publication | 45.3% | 144 of 318 |

## Two examples from the rows

CVE-2026-41709 (Broadcom VMware ESX) was added on 2026-10-28 with an NVD base score of 2.7 and an EPSS of 0.00297 on that day. A CVSS ranking places it near the bottom. CISA listed it on evidence of exploitation.

CVE-2026-9082 (Drupal Core) was added on 2026-05-22 with a base score of 9.8 and an EPSS of 0.00017 on listing day. High on CVSS, near the floor on EPSS, and in KEV two days after NVD published it. Only the KEV flag ranks it correctly on that day.

## The other direction

Severity-only ranking also overfills the top. Of the 123 additions with a CVSS of 9.0 or above and an EPSS score that day, 88 had an EPSS below 0.10. In a backlog that is not limited to KEV entries, the number of unlisted 9.0-and-above CVEs is far larger, and the ledger cannot count them because it contains only listed entries.

That limit matters. The ledger measures how exploited CVEs look on each signal. It does not measure how many unexploited CVEs look the same, so it supports a claim that CVSS misses exploited CVEs, not a claim about precision.

## What to change in the ranking

Put the exploitation evidence first, then use probability for what is not yet listed, then use severity to order within a tier. The guide on building a weighted priority score shows one way to write that down with real CVEs.

Record the 58 additions with no EPSS score on listing day as unscored rather than as zero. Treating a missing value as zero would push them down for a reason unrelated to risk.

## How to repeat this check on your own data

Export your closed and open findings with CVSS, the KEV flag and the EPSS probability on a dated snapshot. Count the KEV-listed findings below 7.0, as the ledger does with its 10.4%. If your share is much lower, check whether your scanner reports vendor scores that differ from NVD.

Then count the KEV-listed findings with an EPSS below 0.10. The ledger found 75.8% of 260 on listing day. Your numbers will differ with the age of your findings, since EPSS rises and falls after listing. Keep the snapshot date with the result.

## A note on the median

The median EPSS on listing day was 0.006555. The median percentile on the same day was 0.55, which places a typical newly listed CVE in the middle of all scored CVEs. Both figures say the same thing from two angles: most KEV entries do not look unusual to the model on the day CISA lists them. That is a limit on EPSS as an early warning for the entries that are listed shortly after publication, and 45.3% were listed within 7 days. It is also a reason to read KEV the day it changes.

## Sources

- FIRST, CVSS v3.1 Specification Document: https://www.first.org/cvss/v3-1/specification-document (read 10 Nov 2026)
- FIRST, CVSS v4.0 Specification Document: https://www.first.org/cvss/v4.0/specification-document (read 10 Nov 2026)
- FIRST, EPSS frequently asked questions: https://www.first.org/epss/faq (read 10 Nov 2026)

Last reviewed 10 Nov 2026. Canonical: https://epssvscvss.com/guides/why-cvss-alone-misranks/
