NUMBERS GUIDE, read 10 Nov 2026
A CVSS vector string is the evidence and the score is only its summary
The v3.1 string, field by field
A v3.1 vector begins with the version label, then lists the base metrics separated by slashes. FIRST's specification names eight: Attack Vector (AV), Attack Complexity (AC), Privileges Required (PR), User Interaction (UI), Scope (S), Confidentiality (C), Integrity (I) and Availability (A).
Take CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H. Read it aloud: reachable over the network, low complexity, no privileges, no user interaction, scope unchanged, high impact on all three properties. Applying the specification's base equations gives 9.8 (Critical). The 9.8 adds nothing the string did not already say.
| Metric | Common values | What it tells the analyst |
|---|---|---|
| AV | N, A, L, P | How far from the target the attacker can be: network, adjacent, local, physical |
| AC | L, H | Whether conditions outside the attacker's control must line up |
| PR | N, L, H | The access the attacker needs before the attack |
| UI | N, R | Whether a person must act |
| S | U, C | Whether the impact crosses a security boundary |
| C / I / A | N, L, H | Damage to confidentiality, integrity and availability |
The severity bands
FIRST's qualitative scale for v3.1 is None at 0.0, Low at 0.1 to 3.9, Medium at 4.0 to 6.9, High at 7.0 to 8.9 and Critical at 9.0 to 10.0. The bands are labels for ranges. A 6.9 and a 7.0 sit in different bands and are separated by one tenth of a point.
Do not use the band as a deadline unless your policy says so. The ledger of KEV additions shows 33 of 318 below 7.0 and 146 at 9.0 or above, so both ends of the scale contain exploited CVEs.
What v4.0 changes
CVSS v4.0 keeps the idea and changes the shape. FIRST states that CVSS consists of four metric groups: Base, Threat, Environmental and Supplemental. Impact is split into the Vulnerable System and the Subsequent System, and a new base metric, Attack Requirements, captures prerequisite deployment and execution conditions.
The specification also asks that a numeric value be labeled by what produced it: CVSS-B for Base only, CVSS-BT, CVSS-BE and CVSS-BTE as groups are added. A bare 9.3 with no label is ambiguous. Ask whether it is CVSS-B or CVSS-BT before comparing it with anything.
The score you receive is usually base only
Scanners and NVD mostly publish base scores. The v4.0 document states that consumers of CVSS should enrich the Base metrics with Threat and Environmental metric values, and that doing so is not required but is highly recommended for more meaningful results.
In practice that means the Environmental group is yours to fill in: the modified base values that reflect your controls and the asset's role. If you do not fill it in, you are ranking by the vendor's view of the flaw in a generic environment.
A reading routine for audits
Record four things for every CVE a decision depends on: the version, the full vector, who scored it (NVD or the vendor) and the date you read it. The ledger takes the NVD base score in the order v3.1, v3.0, v4.0, then v2, with the NVD primary source first, and states that rule in its method.
When two sources give different vectors for one CVE, keep both and name the one you ranked by. Disagreement between NVD and a vendor is common, and an auditor will ask which you chose.
Common misreadings
The first misreading is treating the band as the priority. A Critical label says the base metrics are severe. It does not say the flaw is reachable in your network or that anyone is using it. The second is comparing scores across versions. A v3.1 score of 9.8 and a v4.0 score of 9.3 are produced by different equations from different metrics and should not be sorted in one column without noting the version.
The third is reading Scope or the v4.0 Subsequent System impact as an asset fact. They describe the flaw's reach across a security boundary in a generic deployment. Whether a boundary matters on your asset is an environmental question, and it belongs in your own modified values.
Worked contrast
Two CVEs can share a 9.8 and differ in the string. One has AV:N and PR:N, so any host that can reach the service can attack it. Another has AV:N, PR:L and UI:R, so the attacker needs an account and a person to click. The number matches and the exposure does not. Sort by vector fields when the question is reachability, and by score only when the question is a rough order of severity. In the ledger every one of the 318 KEV rows carries a v3.1, v3.0, v4.0 or v2 score, and the version is recorded per row for this reason.
Sources
- FIRST, CVSS v3.1 Specification Document, read 10 Nov 2026
- FIRST, CVSS v4.0 Specification Document, read 10 Nov 2026
Last reviewed 10 Nov 2026.