What a CVSS score means: CVSS v4.0 metrics, vector strings and severity bands explained
A CVSS number rates a flaw in the abstract, not your risk: how v4.0 builds a score, how to read the vector and how to triage with it.
· 5 min read
Written by AI from the sources linked below. How we use AI
On this page
In short
A CVSS score is a number from 0.0 to 10.0 that rates how severe a software flaw is in itself. It does not say how likely you are to be attacked or how bad it would be for your systems. Most scores you see on advisories are Base scores only, so use them to sort, then check whether the flaw is being exploited and whether it touches something you run.
What CVSS is
The Common Vulnerability Scoring System (CVSS) is a standard maintained by FIRST, the Forum of Incident Response and Security Teams. Its specification says it captures "the principal technical characteristics" of software, hardware and firmware vulnerabilities and outputs numerical scores indicating severity relative to other vulnerabilities. The current version is CVSS v4.0, and you will still see v3.1 scores on older advisories. This guide follows the v4.0 specification.
CVSS is an open framework, so any vendor can score a flaw the same way and anyone can see how the number was derived. That is its main benefit: a "9.8" from two different vendors means the same thing. It is also its limit, as we will see below.
The four metric groups
CVSS v4.0 has four groups of metrics. The first is the one most people see.
- Base: the intrinsic characteristics of the flaw, which do not change over time or between customers. It assumes the reasonable worst case across deployments. The vendor or an analyst working for it normally supplies it.
- Threat: adjusts the score for things like proof-of-concept code or active exploitation. In v4.0 this is one metric, Exploit Maturity (E), with values Not Defined, Attacked, Proof-of-Concept and Unreported.
- Environmental: adjusts the score for your environment: mitigations you have in place and how important the affected system is to you.
- Supplemental: extra context such as whether an attacker can automate exploitation across many targets ("Automatable") and a vendor's own urgency rating. The specification says no supplemental metric changes the final score; it is up to you how to use them.
The specification says that threat and environmental information is only available to the end consumer, which is you. The vendor cannot know whether you have put the flaw behind a firewall.
How the Base score is built
The Base group has two halves.
Exploitability metrics describe how the flaw can be reached:
- Attack Vector (AV): Network, Adjacent, Local or Physical. The more remote an attacker can be, the higher the score. "Network" is what people call remotely exploitable and includes the whole internet.
- Attack Complexity (AC): Low or High, for defences or conditions the attacker has to work around.
- Attack Requirements (AT): None or Present. New in v4.0, it covers conditions that arise from how the system is deployed and run, such as having to win a race condition.
- Privileges Required (PR): None, Low or High.
- User Interaction (UI): None, Passive or Active.
Impact metrics describe the damage, as High, Low or None for confidentiality, integrity and availability. v4.0 scores them twice: once for the vulnerable system (VC, VI, VA) and once for subsequent systems (SC, SI, SA) that suffer downstream. That replaces the single "scope" flag in v3.1.
An important rule in the specification: Base metrics are assessed assuming the attacker knows the target's general configuration and default defences. Target-specific controls such as custom firewall rules belong in the Environmental group, not in the Base score.
Reading a vector string
Every score comes with a vector string that lists the metric choices. It starts with CVSS:4.0 and then lists the metrics in a fixed order, separated by slashes, for example:
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N
This one means: exploitable over the network (AV:N), low complexity, no special deployment conditions, no privileges, no user interaction, high impact on confidentiality, integrity and availability of the vulnerable system, and no downstream impact. All eleven Base metrics must be in the string; Threat, Environmental and Supplemental ones are optional and count as Not Defined when left out. Always read the vector, not only the number: two flaws with a 9.3 can differ greatly in how they can be reached.
The severity bands
The specification maps scores to labels (its Table 22). Using the labels is optional.
A Base score of 5.0, the specification's example, is Medium.
Why the label on the score matters
v4.0 introduces a naming convention so you know what a number contains. CVSS-B is Base only, CVSS-BT adds Threat, CVSS-BE adds Environmental and CVSS-BTE has all three. The specification says assessment providers, and the National Vulnerability Database, typically provide only Base scores, which makes them CVSS-B. A bare "9.8" is therefore almost always a measure of the flaw in the abstract, not of your risk.
Why a high score is not the same as high risk
The specification itself says CVSS is input to a wider process. It lists factors that are outside its scope: regulatory requirements, number of customers affected, monetary losses, life or property threatened and reputational impact.
The scale of the problem is why this matters. A Register report on 13 November 2026 said VulnCheck counted 100,222 CVEs published so far in 2026, and that about 1.4 percent (1,363) had been added to VulnCheck's Known Exploited Vulnerabilities catalog (see our write-up of the 100,000 milestone). The same report quotes VulnCheck's median time from CVE publication to exploitation falling from 120 days in 2025 to 80 days in the first half of 2026. If you tried to patch every "High" or "Critical", you would never finish; if you waited for a 9.0, you could miss a 6.5 that is being exploited this week.
A practical triage routine
- Is it exploited? Check a known-exploited list, such as CISA's catalog or the vendor advisory's own wording. An attacked flaw outranks any theoretical score. (This is what the Exploit Maturity metric records.)
- Is it reachable? Look at Attack Vector and Privileges Required. A network flaw with no privileges needed on an internet-facing system is the top of the list; a local flaw on a laptop with no shared users is not.
- Do you run the affected version? Match the product and version in the advisory against your inventory.
- How much does the system matter? This is the Environmental part: a flaw on a test box and one on the database holding customer records are not equal.
- Is there a mitigation? If a patch cannot go out today, a workaround or a network control can lower the risk until it can.
Common mistakes
- Treating the Base score as risk. It is a measure of the flaw, with the worst case assumed.
- Comparing scores across versions. v3.1 and v4.0 are calculated differently, so a 7.5 and a 7.5 are not guaranteed to match.
- Ignoring the vector. The string tells you what has to be true for an attack to work.
- Forgetting Supplemental metrics. "Automatable" flags flaws that can be wormed through many systems, which is a good reason to patch early even if the score is moderate.
Sources
Written by
The official account of the people who run audiotale.net: announcements, conversation starters and the newsroom's drafts. Posts written with AI say so. Every other account here is a real person.
Read next
What a CVSS score means: how to read a vector string, and why a 10.0 is not the same as risk
A plain guide to CVSS 3.1 and 4.0 bands, metrics and vector strings, with real examples from this week's GitLab patch and how to pair scores with KEV and EPSS.
audiotale.net team7 min read
2026 CVE count passes 100,000: VulnCheck tallies 100,222, but only 1.4% are exploited
AI-driven bug hunting pushed CVE volume past 100,000 this year; VulnCheck's data shows exploitation has not kept pace, and what to prioritise instead.
audiotale.net team7 min read
CVE-2026-116606 and CVE-2026-111560: Barracuda patches root-access chain in ESG, WAF and more
Two Barracuda flaws can be chained to take root on an Email Security Gateway; five other products share the command injection bug. Patch IDs and what to do.
audiotale.net team8 min read