Reading CVE records and CVSS vectors

What a CVE record does and does not contain, who writes it, and how to read CVSS v3.1 and v4.0 vector strings metric by metric, so you can judge a score instead of copying it, and adjust it for your environment.

● Must know for interviewsFREEBy ·Lesson·12 min read·Intermediate → Advanced
CVECVSS v3.1CVSS v4.0NVDvulnerability managementCWE

01What a CVE record is

✓What you will learn
  • Read a CVE record: description, affected versions, references, CWE and scores
  • Decode a CVSS v3.1 vector and see why a score is what it is
  • Read the CVSS v4.0 changes: attack requirements, subsequent systems, threat and environmental groups
  • Know why a base score is not risk, and what to add

A CVE (Common Vulnerabilities and Exposures) ID is a name for one publicly disclosed vulnerability, such as CVE-2024-3400. IDs are assigned by CNAs (CVE Numbering Authorities): vendors like Microsoft or Palo Alto Networks for their own products, plus coordinators such as MITRE and CERT/CC. The CNA writes the record; downstream databases such as the US NVD (National Vulnerability Database) and vendors' own advisories add analysis on top.

Part of the recordWhat it tells youWatch out for
DescriptionThe flaw, the component and the impact in one or two sentencesVague wording such as "may allow" hides whether exploitation is remote or needs authentication
Affected products and versionsWhich versions are vulnerable and where the fix startsVersion ranges are often incomplete early on; the vendor advisory is authoritative
ReferencesVendor advisory, patch, write-ups, sometimes exploit codeThe vendor advisory usually carries workarounds and detection guidance the record lacks
CWEThe weakness class, for example CWE-78 OS command injectionUseful for trends; says nothing about exploitability
CVSSSeverity vectors from the CNA, NVD or CISA's ADP containerDifferent sources can score the same CVE differently; record which one you used
StatePublished, Reserved or RejectedA Reserved ID with no details is common during coordinated disclosure
iNVD is not the only source

Since early 2024 NVD has had a large backlog of unanalysed CVEs, so many records carry only the CNA's score, or none. CISA's Vulnrichment programme adds CVSS, CWE and SSVC data to many records through the ADP container. Build your process so a missing NVD score does not mean a missing priority.

02CVSS v3.1, metric by metric

CVSS (Common Vulnerability Scoring System, maintained by FIRST) scores the severity of a vulnerability from 0.0 to 10.0. The vector string is more useful than the number, because it shows the reasoning. In v3.1 the base metrics describe exploitability (how hard it is to attack) and impact (what an attacker gets).

Five common v3.1 vectors and their base scorestext
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H    -> 9.8 Critical
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H    -> 10.0 Critical
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H    -> 8.8 High
CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H    -> 7.8 High
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:L/I:L/A:N    -> 6.1 Medium (typical reflected XSS)
MetricValuesPlain reading
AV Attack VectorN network, A adjacent, L local, P physicalWhere the attacker must be. N means over the internet if the service is exposed
AC Attack ComplexityL low, H highDoes success depend on conditions outside the attacker's control (a race, a specific configuration)?
PR Privileges RequiredN none, L low, H highDoes the attacker need an account first?
UI User InteractionN none, R requiredDoes a victim have to click or open something?
S ScopeU unchanged, C changedCan the impact reach beyond the vulnerable component, such as a VM escape or XSS reaching the browser?
C / I / AH high, L low, N noneConfidentiality, integrity and availability impact

Severity bands are the same in v3.1 and v4.0: 0.1-3.9 Low, 4.0-6.9 Medium, 7.0-8.9 High, 9.0-10.0 Critical. Read the vector first: AV:N/PR:N/UI:N (remote, unauthenticated, no click) is the shape of mass exploitation, whatever the impact metrics say. AV:L means the attacker already needs code execution on the box, which makes it a privilege-escalation step, not a front door.

!Base score is worst case, not your case

CVSS base scores assume a reasonable worst-case deployment. A 9.8 in a library you ship but never call, or on a server with no network path from anywhere untrusted, is not a 9.8 for you. That is what the temporal or threat metrics and the environmental metrics are for, and most tools ignore them.

03What changed in CVSS v4.0

CVSS v4.0 was published by FIRST in November 2023. It keeps the 0 to 10 scale but restructures the metrics and, importantly, names scores by which groups were used, so nobody mistakes a base score for a risk score.

v4.0 vectors: the same remote unauthenticated full-impact flaw scores 9.3, not 9.8text
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          -> 9.3 (CVSS-B)
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/E:U      -> lower (CVSS-BT): no known exploitation or PoC
CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:P/VC:H/VI:L/VA:N/SC:L/SI:L/SA:N          -> needs attack requirements and passive user interaction
Changev3.1v4.0
Attack conditionsAC folds everything into low or highAC (defences the attacker must bypass) plus new AT Attack Requirements (deployment conditions such as a race or a non-default setting)
User interactionN or RN, P passive (victim just views something), A active (victim must take specific steps)
ScopeS:U or S:CRemoved. Impact is split into the vulnerable system (VC/VI/VA) and subsequent systems (SC/SI/SA)
Temporal / threatExploit maturity, remediation level, report confidenceOnly E Exploit Maturity: Attacked, POC, Unreported
SupplementalNoneSafety, Automatable, Recovery, Value Density, Response Effort, Provider Urgency: context that does not change the score
NamingOne number called the CVSS scoreCVSS-B base, CVSS-BT base and threat, CVSS-BE base and environmental, CVSS-BTE all three

The E metric in v4.0 is the one most worth filling in yourself, because it pulls threat intelligence into the number: E:A (attacked) leaves the score as high as it goes, while E:U (no known exploitation or proof of concept) brings it down. Feed it from the KEV catalogue and EPSS, covered in the next lessons. The environmental group lets you raise or lower the required confidentiality, integrity and availability for the specific asset.

▭On the whiteboard · Reading any vector in ten seconds5 points
01
reach
AV: can it be hit from where attackers are?
02
friction
AC, AT, PR, UI: what stands in the way?
03
impact
C/I/A or VC/VI/VA plus SC/SI/SA: what do they get, and does it spread?
04
threat
E (v4.0): is anyone actually exploiting it?
05
context
Is this asset exposed, critical, compensated by other controls?

04From record to advisory to your estate

The CVE record names the problem; the vendor advisory tells you what to do. For every vulnerability you triage, collect the same few facts so the decision is repeatable.

A triage record: the facts that decide priorityyaml
cve: CVE-2024-3400
product: Palo Alto Networks PAN-OS GlobalProtect
weakness: CWE-77 command injection (via arbitrary file creation)
vector_cna: CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H/E:A   # 10.0 in the vendor advisory
affected: specific PAN-OS 10.2, 11.0 and 11.1 releases with GlobalProtect gateway or portal configured
fixed_in: hotfix releases listed in the vendor advisory
workaround: vendor threat prevention signature (where subscribed); configuration-based mitigation was later withdrawn as ineffective
exploited: yes; listed in CISA KEV in April 2024
our_exposure: 2 internet-facing firewalls run 11.1 with GlobalProtect portal enabled
decision: emergency change, patch within 24h, hunt for indicators from the vendor advisory
  • Configuration matters: CVE-2024-3400 affected firewalls only with GlobalProtect configured. A version match without the vulnerable feature enabled is a lower priority, but confirm the configuration, do not assume it.
  • Workarounds change: vendors revise mitigations as they learn more. Re-read the advisory when you schedule the change, not only when you first triage.
  • Scanners match versions, not configurations: expect false positives (backported fixes on Linux distributions) and false negatives (software installed outside the package manager).
Test yourself

A scanner reports a 9.8 (`AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H`) on a library inside an internal batch application that has no network listener. What is the right next step?

05Putting it into practice

“How to say it in the room

"I read the vector before the number: network-reachable, no privileges, no user interaction is the shape of mass exploitation. Then I check the vendor advisory for affected configurations and workarounds, add threat context, and adjust for the asset. In v4.0 that is the difference between a CVSS-B score from the CNA and a CVSS-BTE score that reflects us."

◎Interview radar

Expect to decode a vector string aloud and explain why a 9.8 might not be your top priority. Knowing that v4.0 dropped Scope for subsequent-system impacts, added Attack Requirements, and labels scores B, BT, BE and BTE shows you have kept up.

Q1Why do CNA and NVD scores differ?+

They score from different information and assumptions. The CNA knows the product; NVD often has only the public description. Record which source you used and prefer the one with the clearer rationale.

Q2What is CWE useful for in triage?+

Mostly trends and root-cause work: if half your findings are CWE-79 or CWE-89, invest in framework-level fixes. It does not change the priority of one finding.

Q3What does a Rejected CVE mean?+

The ID was withdrawn: a duplicate, not a vulnerability, or never used. Scanners occasionally keep reporting rejected IDs; close them with a note.

Finished this one?
Tick it and it shows as done on the roadmap and the course.