← All posts
Legislation & compliance

CVSS explained: how severe is a vulnerability?

You open your pentest report. Finding 1: Critical, 9.8. Finding 2: Medium, 6.1. Finding 3: Low, 2.5.

Clear enough: start with the 9.8. But where do those numbers come from? Did the pentester just pick them? And is a 9.8 always worse than a 6.1?

The numbers come from CVSS, the Common Vulnerability Scoring System. In this blog I'll explain what CVSS is, why we use it, how a score is built up and, just as important, what a score does not tell you.

Want to try it while reading? Open our free CVSS calculator in another tab.

What is CVSS?#

CVSS is an open standard for describing how severe a software vulnerability is. It is free to use and is owned and managed by FIRST (Forum of Incident Response and Security Teams), a US-based non-profit organization that supports security incident response teams around the world[1].

CVSS has been around since 2005 and has been revised several times. Version 3.0 appeared in 2015, version 3.1 in 2019, and since November 2023 there is also version 4.0[1].

The result of CVSS is always two things:

  1. A score from 0.0 to 10.0.
  2. A vector: a short line of text that shows exactly how that score was reached, for example CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H.

That vector is the real strength of CVSS. Anyone can take it, put it into a calculator and check whether they agree with every choice that was made.

Why do we use CVSS?#

Imagine three pentesters assessing the same vulnerability. Without a shared system, one calls it "serious", the second "pretty bad" and the third "high risk". Which one is right?

CVSS solves that with a fixed set of questions. That gives you:

  • One common language. A 7.5 means the same to your IT supplier, your software vendor and us.
  • Comparable findings. Findings from different tests, or from different years, can be put side by side.
  • Clear priorities. You know what to fix first, and you can explain to management why.
  • Transparency. Thanks to the vector, every score can be checked. No black box.

That's also why public vulnerability databases use CVSS. The best known is the American National Vulnerability Database (NVD) of NIST, which publishes a CVSS score for known vulnerabilities (CVEs)[2]. When you read in the news that a vulnerability "scores 9.8", that is almost always a CVSS score.

How is the score calculated?#

CVSS 3.1 has three metric groups[3]:

  • Base: the properties of the vulnerability itself, which don't change over time or between organizations.
  • Temporal: properties that do change over time, for example whether there is already a working exploit or a patch.
  • Environmental: properties that depend on your own environment, such as how important the affected system is to you.

When people say "the CVSS score", they almost always mean the base score. That's also what our calculator calculates. It answers two questions: how easy is it to exploit this vulnerability? and how much damage can it do?

Part 1: how easy is the attack?#

Attack Vector (AV): where does the attacker need to be?

ValueMeaning
NetworkAnywhere on the internet. The worst case.
AdjacentOn the same local network, for example the same Wi-Fi or office network.
LocalAlready on the system, for example logged in or via a file the user opens.
PhysicalPhysically at the device, for example with a USB stick.

Attack Complexity (AC): does the attack need special conditions? Low means the attack works every time. High means the attacker depends on circumstances outside their control, such as a specific configuration or winning a race against the system.

Privileges Required (PR): does the attacker need an account? None means an anonymous visitor can do it. Low means a normal user account is needed. High means administrator rights are needed.

User Interaction (UI): does a victim need to do something? None means the attack works fully on its own. Required means someone has to click a link or open a file first.

Part 2: how far does the damage reach?#

Scope (S): does the attack stay within the vulnerable component? This one sounds abstract, so here's an example. A vulnerability in a website that only affects that website is Unchanged. A cross-site scripting (XSS) vulnerability in a website that runs code in the visitor's browser is Changed: the vulnerability is in the website, but the damage happens somewhere else. When the scope is changed, the score goes up.

Part 3: what is the impact?#

The last three metrics are the classic trio of information security: Confidentiality, Integrity and Availability. Each one is High, Low or None.

MetricQuestionHigh means
Confidentiality (C)Can the attacker read data they shouldn't?All data can be read, or very sensitive data like passwords.
Integrity (I)Can the attacker change data?The attacker can change anything they want.
Availability (A)Can the attacker disrupt the system?The system can be taken down completely.

From score to label#

The score is then translated into a label with FIRST's qualitative severity rating scale[4]:

ScoreSeverity
0.0None
0.1 - 3.9Low
4.0 - 6.9Medium
7.0 - 8.9High
9.0 - 10.0Critical

Examples from practice#

Theory is fine, but CVSS really comes to life with examples. Click a vector to open it in our calculator.

Log4Shell: 10.0 (Critical)#

In December 2021, a vulnerability in Log4j, a logging library used in countless Java applications, put the whole internet on edge. NVD gave it the maximum score[5]:

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H

Read it from left to right: from the internet (AV:N), no special conditions (AC:L), without an account (PR:N), without anyone clicking anything (UI:N), with impact beyond the vulnerable component (S:C) and full control over confidentiality, integrity and availability (C:H/I:H/A:H). There is literally nothing that makes this attack harder.

Heartbleed: 7.5 (High)#

Heartbleed (2014) was a bug in OpenSSL that allowed attackers to read pieces of a server's memory, including passwords and private keys[6]:

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N

Just as easy to exploit as Log4Shell, but the attacker can only read (C:H), not change anything or take the server down (I:N/A:N). That's why it ends up at 7.5. Still serious: stolen private keys could be used to intercept "secure" connections.

Reflected XSS: 6.1 (Medium)#

A classic web application finding, like the XSS vulnerability we found at the municipality of Eindhoven. A typical reflected XSS scores:

CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:L/A:N

The victim has to click a prepared link (UI:R), and the attacker can then read and change a limited amount of data in the victim's browser (C:L/I:L). The scope is changed, because the vulnerable website is not the same as the victim's browser where the code runs.

A hard-to-exploit local leak: 2.5 (Low)#

CVSS:3.1/AV:L/AC:H/PR:N/UI:R/S:U/C:L/I:N/A:N

The attacker already has to be on the system (AV:L), needs special circumstances (AC:H) and the help of a user (UI:R), only to read a little bit of information (C:L). Worth fixing, but it won't keep anyone up at night.

A score is not a risk#

This is the most important part of this blog. FIRST itself emphasizes it: CVSS measures the severity of a vulnerability, and is not meant to be used on its own to assess risk[3].

The difference? Take two findings:

  • A 9.8 on a test server that is switched off, contains no data and can't be reached from the internet.
  • A 6.1 on the login page of your webshop, which thousands of customers use every day.

On paper, the 9.8 is worse. In practice, the 6.1 is probably the bigger risk to your business. CVSS can't know that, because CVSS doesn't know your business.

That's why a good pentest report doesn't stop at a number. In our reports, every finding gets a CVSS vector and score, so you can compare findings objectively. But we also explain what the finding means for your organization, and in which order it makes sense to fix things.

What about CVSS 4.0?#

Since November 2023 there is CVSS 4.0[1]. It has four metric groups (Base, Threat, Environmental and Supplemental), new metrics, and makes it even more explicit that a base score alone is not a risk score[7]. You'll see v4.0 scores more and more, for example in our blog about PHP2Shell, where the PHP team rated the vulnerabilities at 9.2 on CVSS v4.0.

Vulnerable system and subsequent systems#

The biggest change in the base metrics: the abstract Scope metric is gone. Instead, CVSS 4.0 scores the impact twice[7]:

  • Vulnerable System Impact (VC, VI, VA): the damage to the system that contains the vulnerability itself.
  • Subsequent System Impact (SC, SI, SA): the damage to other systems that are affected as a result, such as a visitor's browser, a database behind the application or other machines on the network.

Take the reflected XSS from earlier. The website contains the vulnerability, but the attacker's code runs in the victim's browser. In v3.1, that was "Scope: Changed". In v4.0, you score it directly: no impact on the website itself, some impact on the browser that comes after it:

CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:P/VC:N/VI:N/VA:N/SC:L/SI:L/SA:N

Two other changes you'll see in the calculator: the new metric Attack Requirements (AT), for conditions in the target's own setup that the attack depends on, such as a specific configuration or a position between the victim and the server. And User Interaction is now split into Passive (the victim only has to visit a page or view something) and Active (the victim has to click, open a file or accept a warning).

CVSS 3.1 is still very widely used, though: many advisories, tools and reports still give a 3.1 score. That's why our calculator does both: use the switch at the top to choose between 3.1 and 4.0. For 4.0 it calculates the base score, which FIRST calls CVSS-B. The severity labels (Low to Critical) have stayed the same in v4.0[7], so the scale above still applies.

Conclusion#

CVSS gives vulnerabilities a common language: a score from 0 to 10, a label and a vector that shows exactly how it was reached. That makes findings comparable and helps you prioritize.

But remember: a score describes the vulnerability, not your risk. Use CVSS as a starting point, and then look at what's actually at stake for your organization.

Want to calculate a score yourself? Use our CVSS calculator. And want to know which vulnerabilities are hiding in your own systems? That's exactly what a pentest is for.

Sources#

[1] https://www.first.org/cvss/
[2] https://nvd.nist.gov/vuln-metrics/cvss
[3] https://www.first.org/cvss/v3.1/user-guide
[4] https://www.first.org/cvss/v3.1/specification-document
[5] https://nvd.nist.gov/vuln/detail/CVE-2021-44228
[6] https://nvd.nist.gov/vuln/detail/CVE-2014-0160
[7] https://www.first.org/cvss/v4.0/user-guide

Melis IT Solutions
Jeremy Melis19-11-2026