Skip to content

CVE vs CWE vs CVSS vs CPE: How They Fit Together

CVE vs CWE vs CVSS vs CPE: How They Fit Together

What Is the Difference Between CVE, CWE, CVSS and CPE?

CVE, CWE, CVSS and CPE are four separate standards that describe one security flaw from four angles. CVE gives the flaw a unique ID, CWE names the type of mistake behind it, CVSS rates how severe it is, and CPE names the products it touches. Vulnerability databases stack all four on one record, yet each answers a different question.

Mixing them up causes real mistakes. The CVE vs CWE mix-up is common, and so is treating a CVSS 10.0 as proof of an attack. Below, one real flaw (Log4Shell) walks through all four, and the same logic feeds a vulnerability management plan, CVSS 4.0 triage of high and medium CVEs and our CVSS calculators.

One flaw, four questions Four rows pair a question with the identifier that answers it: which flaw is it, CVE; what mistake caused it, CWE; how severe is it, CVSS; which products and versions, CPE. One flaw, four questions Which flaw is it?CVE What mistake caused it?CWE How severe is it?CVSS Which products and versions?CPE
Figure 1. The four identifiers each answer one question about the same flaw. Conceptual diagram. Source: UnpanicTech original.

Key Takeaways

  • A CVE names one specific flaw in one product. It does not grade severity.
  • A CWE names the kind of mistake, and one CWE can sit behind many CVEs.
  • CVSS rates severity from 0.0 to 10.0. It does not measure your risk.
  • CPE is a naming scheme for products, and NVD uses it to say what a CVE affects.
  • Log4Shell (CVE-2021-44228) shows all four, plus a CISA KEV listing.
  • Severity and exploitation are different facts. See how CISA KEV additions work in practice.

CVE, CWE, CVSS and CPE come from different maintainers, which is why they behave differently in practice. MITRE describes CVE as the de facto global standard for identifying vulnerabilities, and CWE as a common language for the weaknesses behind them [2]. FIRST owns and manages CVSS, an open framework that rates severity and is free to use with attribution [3]. NIST's National Vulnerability Database hosts the official CPE dictionary, a naming scheme for IT systems, software and packages [4].

NVD sits downstream of the CVE Program and has historically added CWE mappings, CVSS and CPE information to CVE records [2]. So the CVE is the shared name, and the other three are layers that different groups attach to it. When two tools disagree about a flaw, find out which layer disagrees and who supplied it.

What Do CVE, CWE, CVSS and CPE Each Tell You?

CVE, CWE, CVSS and CPE each answer one question. CVE identifies a specific vulnerability in a specific product. CWE classifies the underlying weakness, such as deserialization of untrusted data. CVSS rates severity on a scale from 0.0 to 10.0. CPE gives products and versions a standard name so tools can match a flaw to what you run. None of the four replaces another.

Definition: CVE ID

A CVE ID is the shared name for one specific vulnerability in a specific product. Other programs attach more data to that name: MITRE's CWE adds the weakness type [2], FIRST's CVSS adds a severity score [3], and NVD adds CPE matches [4].

MITRE separates a weakness from a vulnerability this way: the weakness is the general condition, and the vulnerability is one concrete instance of it in one product [2]. That is why one CWE entry can be the root cause behind many different CVE records [2]. Our vulnerability management guide shows where each field lands in a ticket.

The table below compares CVE, CWE, CVSS and CPE side by side, including what each one cannot tell you.

StandardQuestion it answersMaintained byWhat it cannot tell you
CVEWhich flaw is this?CVE ProgramHow bad it is
CWEWhat kind of mistake caused it?MITREWhich single product is affected
CVSSHow severe is it?FIRSTWhether attackers use it
CPEWhich product and version?NIST (NVD dictionary)How the flaw works

Common mix-up: CWE is not a smaller CVE. A CWE describes a class of mistake, while a CVE points at one flaw in one product.

Three lines worth remembering

  • A CVE tells you which flaw. A CWE tells you what kind of mistake caused it.
  • CVSS scores severity in a vacuum. It never knows whether your server is exposed.
  • Only exploitation evidence tells you attackers have already used a flaw.

Can You See All Four on One Real Vulnerability?

Yes. Log4Shell, tracked as CVE-2021-44228, carries all four identifiers. The CVE ID names the flaw in Apache Log4j2, CWE entries such as CWE-502 classify the mistake, and CVSS 3.1 rates it 10.0 Critical. CPE names are how tools match it to a product and version. CISA also lists it as a known exploited vulnerability, which CVSS cannot supply by itself.

The table below shows CVE-2021-44228 across the four identifiers, with the source for each field.

FieldValue for CVE-2021-44228
CVE ID and productCVE-2021-44228, Apache Log4j2, component log4j-core [1]
CWECISA's catalog row lists CWE-20, CWE-400 and CWE-502 [5]
CVSS3.1 base score 10.0 Critical, vector CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H [1]
Affected versions2.0-beta9 up to 2.3.1, 2.4 up to 2.12.2, and 2.13.0 up to 2.15.0, each range ending before its upper bound [1]
Fixed versions2.3.1 (Java 6), 2.12.2 (Java 7), 2.15.0 (Java 8 and later) [1]
CPEFollows the CPE 2.3 pattern, for example cpe:2.3:a:apache:log4j:2.14.1:*:*:*:*:*:*:*. This string is illustrative; check NVD for the real match data [4]
KEV statusListed by CISA, added 2021-12-10, ransomware use marked Known (checked November 17, 2026) [5]
Log4Shell across the four identifiers A top box names CVE-2021-44228 in Apache Log4j2. Four boxes below list its CWE, CVSS score, CPE role and CISA KEV listing. CVE-2021-44228 Apache Log4j2, component log4j-core CWECWE-502 (CISA also lists 20, 400) CVSS3.1 base score 10.0, Critical CPENames product and version for matching KEVListed by CISA, added 2021-12-10
Figure 2. How the four identifiers, plus KEV, attach to CVE-2021-44228. Source: UnpanicTech original diagram, built from the Apache and CISA entries cited above.

One detail lives in the advisory text and not in the ID. Apache states that only the log4j-core JAR is affected, and that applications using only log4j-api are not [1]. In our reading, a scanner that matches on a product name alone may flag an application that ships only log4j-api, so the advisory still needs a human read.

The same product also shows how one name hides many CVEs. Apache's page lists these related entries, each with its own score [1]:

  • CVE-2021-44228: 10.0 Critical, code execution through JNDI lookups.
  • CVE-2021-45046: 9.0 Critical, because the 2.15.0 fix was incomplete in certain non-default configurations.
  • CVE-2021-45105: 5.9 Medium, a denial of service from infinite recursion in lookup evaluation.

How Does CVSS Turn a Flaw Into a Number?

CVSS turns a flaw into a score by rating a fixed set of metrics, such as attack vector, complexity, privileges required and impact. Those ratings become a vector string and a number from 0.0 to 10.0. The Base score describes the flaw itself. Threat and Environmental metrics let each organization adjust that score for exploit maturity and local controls.

CVSS is an open framework owned and managed by FIRST, with four metric groups: Base, Threat, Environmental and Supplemental [3]. Base metrics capture traits that stay constant over time and across environments, and they assume a reasonable worst case [3]. Threat metrics adjust for factors such as public proof of concept code or reported attacks, while Environmental metrics adjust for the controls and importance of the system in a given network [3].

Supplemental metrics add context but never change the score [3]. FIRST asks that scores carry a label showing which metrics were used, such as CVSS-B for Base only, and notes that providers like NVD typically publish only Base scores [3]. In practice, the Base score travels with the CVE record, while Threat and Environmental values have to come from the team that runs the system [3]. A published CVSS number is therefore a severity rating, and risk to your network needs more input.

The graphic below breaks the CVSS 3.1 vector for CVE-2021-44228 into its eight metrics, using the vector Apache publishes [1].

CVSS 3.1 vector for CVE-2021-44228 Eight metric cells: network attack vector, low complexity, no privileges, no user interaction, scope changed, and high confidentiality, integrity and availability impact. The result is a base score of 10.0, rated Critical. CVSS 3.1 vector, Log4Shell CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H AV:N Network AC:L Low complexity PR:N No privileges UI:N No user action S:C Scope changed C:H Confidentiality high I:H Integrity high A:H Availability high Base score 10.0 = Critical
Figure 3. Reading the CVSS 3.1 vector for CVE-2021-44228. Source: UnpanicTech original diagram.

FIRST maps scores to a qualitative scale, shown below for CVSS 4.0 [3].

RatingCVSS score
None0.0
Low0.1 to 3.9
Medium4.0 to 6.9
High7.0 to 8.9
Critical9.0 to 10.0

Versions matter. Version 4.0 replaced the old Scope idea with separate impact ratings for the vulnerable system and subsequent systems, and includes an Attack Requirements metric [3]. Apache's page already mixes both worlds: Log4Shell carries a 3.1 score, while newer entries such as CVE-2026-34477 carry a 4.0 vector and a 6.3 Medium score [1]. Never rank a 3.1 number against a 4.0 number as if they were the same measurement.

Labels can also slip. Apache lists CVE-2018-1285 as "9.8 HIGH", although the scale in FIRST's 4.0 specification puts 9.0 to 10.0 in Critical [1][3]. Read the number and the vector, then decide.

Does a High CVSS Score Mean Attackers Are Using the Flaw?

No. A high CVSS score says a flaw would be severe if exploited, not that anyone is exploiting it. Exploitation evidence comes from separate sources, such as CISA's Known Exploited Vulnerabilities catalog, which lists CVEs based on evidence of active exploitation. CVSS can absorb that evidence through its Threat metric group, but only if somebody enters it.

CISA created the KEV catalog under Binding Operational Directive 22-01 as a living list of known CVEs that carry significant risk to the federal enterprise [5]. The directive binds only federal civilian agencies, but CISA urges all organizations to prioritize catalog items in their own vulnerability management [5].

For Log4Shell, the catalog row shows a date added of 2021-12-10 and a due date of 2021-12-24 [5]. A missing KEV entry, in our view, only means CISA has not listed the flaw. It is not a safety certificate.

CVSS 4.0 keeps proof of concept and real attacks apart in its Exploit Maturity metric. The table below shows the three defined values for CVSS exploit maturity [3].

ValueWhat the threat intelligence shows
Attacked (A)Attacks have been reported, or exploit toolkits are available
Proof-of-Concept (P)Public proof of concept code, with no reported attacks
Unreported (U)No known public proof of concept and no reported attacks

How Should a Patch Team Use All Four Together?

A patch team should treat CVE, CWE, CVSS and CPE as a sequence of filters. Match the CVE to your inventory by product and version, read the CVSS vector for exposure conditions, check whether CISA lists the CVE as exploited, then apply the newest vendor fix. The CWE tells you what kind of fix to expect. Severity alone never sets the order.

  1. Match. Take the CVE ID, read the vendor advisory for the affected component and versions, and compare them with your inventory. CPE names help tools do this at scale [4].
  2. Read exposure. In the vector, AV:N with PR:N means the flaw can be reached over a network without credentials [3].
  3. Check exploitation. Look the CVE up in the KEV catalog and write down the date you checked [5].
  4. Patch and record. Move to the newest fixed release in your version line, then log CVE, CWE, score with its CVSS version, and KEV date in the ticket.
Four-step triage using all four identifiers Four stacked steps with arrows between them: match the CVE and CPE to inventory, read the CVSS vector for exposure, check CISA KEV for exploitation, then patch to the newest fixed release and record the result. 1 MatchCVE + CPE + advisory versions 2 Read exposureCVSS vector: AV, PR, UI 3 Check exploitationCISA KEV, with the date you checked 4 Patch and recordNewest fixed release in your line
Figure 4. A triage order that uses CVE, CPE, CVSS and KEV in turn. Conceptual workflow. Source: UnpanicTech original diagram.

Patching Log4Shell shows why a CVE ID is a starting point and not a finish line. Apache's advisory lists Log4j 2.15.0 as the fixed version for CVE-2021-44228 on Java 8 and later [1]. The same page records that the 2.15.0 fix was incomplete in certain non-default configurations, which produced CVE-2021-45046 and a fix in 2.16.0 [1].

Entries for CVE-2021-45105 and CVE-2021-44832 point to 2.17.0 for Java 8 and later [1]. Apache's page also lists three fixed versions for the first CVE alone, one for each Java line it still covers [1]. CISA's catalog lists CVE-2021-44228 with a federal due date of 2021-12-24 [5]. The practical rule for defenders is to read the whole advisory page for the product, not only the one CVE that triggered the alert, and to move to the newest fixed release in the Java line they run.

What Are the Strategic Implications for Your Patch Process?

Your patch process should store CVE, CWE, CVSS, CPE and KEV status as separate fields, because each has a different owner and a different shelf life. Sorting a backlog by one blended number hides that. The CVE and CPE data identify the asset, the CVSS Base score rates the flaw, and exploitation evidence changes the queue.

In our view, the most useful change is ownership. Vendors supply the CVSS Base score, so your team owns the Threat and Environmental overlay that FIRST says only the consumer can set [3]. Record the CVSS version next to every score, since a 3.1 value and a 4.0 value do not compare.

The CWE field pays off later. MITRE describes root cause mapping as a way to see which weakness types keep producing vulnerabilities [2]. If the same CWE keeps appearing in your vendors' CVEs, that is a question for procurement and secure development reviews, not only for the patch team.

Sources & References

Sources checked on November 17, 2026.

  1. Apache Logging Services, "Security" (known vulnerabilities page for Log4j)
  2. MITRE CWE, "Frequently Asked Questions (FAQ)"
  3. FIRST, "Common Vulnerability Scoring System version 4.0: Specification Document"
  4. NIST National Vulnerability Database, "Official Common Platform Enumeration (CPE) Dictionary"
  5. CISA: Known Exploited Vulnerabilities Catalog (print view) and "CISA Adds 95 Known Exploited Vulnerabilities to Catalog"
NK

Cybersecurity Researcher & Technical Editor

UnpanicTech publishes independent cybersecurity research and defensive security analysis. Naseem Khan serves as Technical Editor, with a focus on vulnerability analysis, defensive security, incident response, cloud security, and practical security engineering.

Leave a Comment (Authenticated Users)