SCORING SYSTEM, read 10 Nov 2026
CVSS v4.0 adds a Threat group and sharper nomenclature, but the score the NVD shows is still mostly the Base score
- Maintained by
- FIRST, through the CVSS Special Interest Group
- Scale
- 0.0 to 10.0; same qualitative bands as v3.1 (Low 0.1 to 3.9 up to Critical 9.0 to 10.0)
- Current version
- v4.0 (v1.2 of the specification document, 2024-06-18)
- Updated
- Specification versions 1.0 on 2023-11-01, 1.1 on 2023-11-09, 1.2 on 2024-06-18
- Answers
- The severity of a flaw, adjusted for current threat evidence
What changed from v3.1
The v4.0 specification states that CVSS consists of four metric groups: Base, Threat, Environmental and Supplemental. The Base group represents intrinsic qualities that are constant over time and across environments. The Threat group reflects characteristics that change over time. The Environmental group captures what is unique to a user's environment. The Supplemental group adds extrinsic information that does not change the score.
The Base group gains Attack Requirements (AT), which captures prerequisite deployment and execution conditions, and replaces v3.1 Scope with separate impact metrics for the vulnerable system (VC, VI, VA) and for subsequent systems (SC, SI, SA). User Interaction becomes None, Passive or Active.
The Supplemental group is worth separate mention, because it does not move the score. The specification lists Automatable, Recovery, Value Density, Vulnerability Response Effort and Provider Urgency among its metrics, and says consumers may give them more, less or no effect on categorization. Automatable asks whether an attacker can automate exploitation events across multiple targets, and it is the same concept that the CISA SSVC tree uses as a decision point.
How the score is computed
v4.0 has no closed formula like the v3.1 one. The specification describes scoring using MacroVectors and interpolation: the 15 million CVSS-BTE vectors are gathered into 270 disjoint equivalence sets of comparable qualitative severity, experts compared vectors from each set, and the comparison data was used to order the vectors from least to most severe. A lookup then returns a score for the MacroVector, and interpolation adjusts within it.
The practical consequence is that an analyst cannot reproduce a v4.0 score by hand from a short formula. A calculator or a published implementation is required. This is a real difference for an auditor, who in v3.1 could check the arithmetic with the constants in Table 16.
The Base Score documented by FIRST assumes the highest severity for the Threat and Environmental metrics when they are not set. A score with only Base metrics is therefore CVSS-B, a worst-case figure, and the specification asks that it be labeled that way.
The Threat group and the nomenclature
The Threat group has one metric, Exploit Maturity (E), described as based on exploit techniques, exploit code availability or active in-the-wild exploitation. When it is not defined the score assumes the highest severity, so a missing value does not lower anything.
The specification also requires labeling: because a numerical score means different things depending on the metrics used, scores should carry nomenclature. CVSS-B is Base only, CVSS-BT is Base and Threat, CVSS-BE is Base and Environmental, and CVSS-BTE is all three. A score printed without that label cannot be compared with another.
For a program, the nomenclature is the main practical change. A report column headed CVSS has no meaning in v4.0 unless it says which of the four labels applies. When the NVD or a CNA publishes only Base metrics, the honest label is CVSS-B 4.0.
| Label | Metrics used |
|---|---|
| CVSS-B | Base |
| CVSS-BT | Base and Threat |
| CVSS-BE | Base and Environmental |
| CVSS-BTE | Base, Threat and Environmental |
What the NVD publishes
The NVD news page records an April 2026 incident. On 2026-04-28 NIST announced that about 4,500 CVE records, 19 percent of the records with a v4.0 score, carried a numerical score higher than the correctly calculated value, and fewer than 30 carried a lower one. Only the numerical scores were affected. NIST said it would run a correction script on 2026-04-28 at 9:00 pm EDT and update the modified date on each record.
The same page records the NVD policy change of 2026-04-15: the NVD will no longer routinely provide its own severity score for CVEs where the CNA has already provided one. A v4.0 number in a feed may therefore come from the CNA, not NIST. A program that ingests scores should store the source field next to the value.
Reproduction matters for an audit. The specification points to Section 2.5 of the User Guide for the MacroVector approach, and FIRST provides a calculator and a JSON and XML data representation for exchanging scores. Check that a tool's v4.0 output matches the FIRST calculator on a sample of vectors before trusting a bulk import.
What the KEV Lag Ledger shows
In the ledger of 318 KEV additions (2025-11-10 to 2026-11-09), the CVSS field is v3.1 for 313 rows and v3.0 for 5. None of the 318 selects a v4.0 score under the ledger rule of v3.1, then v3.0, then v4.0, then v2. That rule was chosen so that a single comparable number exists for every row, and it means this site's severity statistics describe the v3 family.
For a program that has moved its policy to v4.0, the check is to rerun the same cut: how many exploited vulnerabilities fall below your v4.0 cutoff. The method is in the ledger introduction. The v3 result of 10.4 percent below 7.0 is a starting point, not a measurement of v4.0.
Using it in a ranking
Cite the version and the nomenclature in policy: for example, CVSS-BT 4.0 with the vector string. Use Threat only when a named, dated source sets Exploit Maturity, such as KEV membership. Treat Supplemental values (Automatable, Safety, Provider Urgency) as context for a human reviewer.
Do not mix v3.1 and v4.0 scores in one sorted list. The scales and the qualitative bands are the same, but the scoring methods differ, so a single vulnerability can land in a different band under each.
Adopt v4.0 when the sources you rely on publish it. Where a feed carries both versions, record both and name the one the policy uses. Re-check the NVD source field before importing, given the April 2026 correction.
Comparisons with this system
Sources
- FIRST, CVSS v4.0 specification document (metric groups, nomenclature, MacroVectors, version history), read 10 Nov 2026
- NVD news: v4.0 score correction (2026-04-28) and enrichment policy (2026-04-15), read 10 Nov 2026
- FIRST, CVSS v3.1 specification document, read 10 Nov 2026
Last reviewed 10 Nov 2026.