What CVSS is
CVSS, the Common Vulnerability Scoring System, is a scoring method for security flaws. It is maintained by FIRST, a global forum of security response teams. A flaw gets a number from 0.0 to 10.0, and higher means more severe.
The score lets a security team sort a long list of problems without reading every report. A 9.8 goes to the top of the pile. A 3.1 can wait for the next maintenance window.
The severity bands
The standard groups scores into ratings:
- None: 0.0
- Low: 0.1 to 3.9
- Medium: 4.0 to 6.9
- High: 7.0 to 8.9
- Critical: 9.0 to 10.0
What goes into a score
CVSS looks at how a flaw can be attacked and what happens if it is. The questions are practical. Can someone exploit it over the internet, or only from the same machine? Does it need a login? Does the victim have to click something? And if it works, can the attacker read data, change it or knock the system over?
A flaw that can be hit remotely, with no credentials and no user action, that leads to remote code execution, lands near the top. The same bug that needs local access and a logged-in user scores far lower. Version 4.0 of the standard, released in 2023, added finer detail such as whether a successful attack affects other systems and whether the flaw is already being exploited.
What a score does not tell you
A CVSS score measures the flaw in general, not your situation. A critical bug in software you do not run is no risk to you. A medium bug in your most exposed server may be your biggest problem.
The score also does not say whether anyone is actually using the flaw. For that you need threat intelligence about active exploits, and you need to know how fast you patch. A high score plus a wide patch gap is the combination to fear.
Why it matters when you use AI
When AI tools find and test bugs quickly, more flaws arrive with scores attached, and the pile grows. Use CVSS as a first sort, then ask three plain questions: do we run this, is it reachable from outside, and is anyone attacking it yet?
Every scored flaw usually has a CVE ID. A zero-day vulnerability is one attackers use before a fix exists, and a buffer overflow is one common type of flaw that scores high.