When the CVE record, the NVD entry and the EUVD entry for the same vulnerability disagree on severity, affected versions or status, most vulnerability pipelines silently pick whichever arrived last. That choice decides patch deadlines, scanner findings and audit evidence, so it should be a written rule, not an accident of ingestion order. This guide compares the three public sources by role, data model, timeliness, governance and failure modes, then gives a source-of-truth policy, field-by-field precedence rules and a reconciliation workflow you can build into a vulnerability management platform or scanner feed. It is for vulnerability management leads, tooling engineers, product security teams and EU-regulated organisations. History of the NVD backlog and the CVE funding episode is covered in the CVE Programme and NVD Backlog Guide, and EUVD features in the EU Vulnerability Database (EUVD) guide.
Status note: source behaviour below was checked on the NIST NVD page, the CISA Vulnrichment repository, the CVE Record Format schema, ENISA announcements, the EUVD public API and the NIS2 text on 14 November 2026.
Key takeaways
- They are three layers, not three rivals. The CVE Programme assigns identifiers and publishes the CNA's record; the NVD enriches a prioritised subset; the EUVD aggregates CVE, advisory, CSIRT and exploitation data under NIS2 Article 12.
- The CNA record is the default source of truth. It is the only layer written by the party closest to the product, and CISA's Vulnrichment rules already give CNA data precedence over its own ADP enrichment.
- The NVD is now selective. Since 15 April 2026 it prioritises CISA KEV entries (goal: one business day), software used in federal government and EO 14028 critical software; the rest are labelled "Lowest Priority - not scheduled for immediate enrichment".
- The EUVD is a join point, not a replacement. Each EUVD record carries its own EUVD-YYYY-NNNNNN identifier with CVE aliases, a score, EPSS and an exploitedSince date, which makes it useful for triage and cross-checking.
- Record provenance per field. A pipeline that stores which source supplied each score, CWE and version range can explain any decision to an auditor and switch sources when one degrades.
The short answer
Treat the CVE record (CNA container first, then ADP containers such as CISA's) as primary for identity, description and affected versions. Use the vendor's own advisory as the tie-breaker for affected versions and fixes. Use NVD enrichment where it exists, mainly for CPE applicability and for KEV-listed CVEs, and record when it is absent. Use the EUVD as an independent cross-check and, for EU entities, as the European reference point for exploited and EU-coordinated vulnerabilities. Exploitation status comes from the CISA KEV catalogue and the EUVD exploited view, never from a CVSS score.
Roles compared
| Dimension | CVE Programme | NVD (NIST) | EUVD (ENISA) |
|---|---|---|---|
| Core function | Assigns CVE IDs and publishes records written by CVE Numbering Authorities (CNAs); Authorised Data Publishers (ADPs) add containers | Enriches CVE records with analysis such as product applicability and severity, now for a prioritised subset | Aggregates CVE data, vendor advisories, national CSIRT alerts and exploited markings into one EU service |
| Legal or governance basis | Programme rules; record format governed by the CVE Quality Working Group with CVE Board final approval | NIST programme; prioritisation set by NIST (15 April 2026 update) | NIS2 Article 12(2): ENISA "shall develop and maintain" the database after consulting the Cooperation Group |
| Identifier | CVE-YYYY-NNNN | Uses the CVE ID | EUVD-YYYY-NNNNNN, with CVE and other aliases |
| Who writes the data | The CNA for the product (vendor, project, CERT or CNA of last resort) | NIST analysts | Automated aggregation plus ENISA and EU CSIRT input; ENISA is a CNA and, since November 2025, a CVE Root for European entities |
| Timeliness | Published when the CNA publishes | KEV goal one business day; others may wait or stay unenriched | Depends on upstream sources; operational since 13 May 2025 |
| Main access | CVE Services API and the cvelistV5 repository | NVD website and API | Web dashboards and public JSON API |
Data fields compared
The CVE Record Format is a JSON schema; any 5.x dataVersion is valid and the current default is 5.2.0. A record is either PUBLISHED or REJECTED. The cna container must carry providerMetadata, descriptions, affected and references, and may carry problemTypes (cweId in the form CWE-n), metrics (cvssV4_0, cvssV3_1, cvssV3_0, cvssV2_0 or other) and cpeApplicability with CPE 2.3 match criteria. Each adp container must identify the publisher that supplied it.
| Field | CVE record | NVD | EUVD (API record) |
|---|---|---|---|
| Description | descriptions (CNA, required) | Reproduces the CVE description | description |
| Affected products and versions | affected (CNA, required): vendor and product or package, with versions or default status | CPE applicability where analysed | Vendor identifier (enisaIdVendor) in the exploited-view record |
| Severity | metrics, optional, by CNA or ADP | NVD score where it chooses to add one; since April 2026 not routinely added when the CNA supplied a score | baseScore, baseScoreVersion, baseScoreVector |
| Weakness | problemTypes cweId, optional | CWE where analysed | Not in the exploited-view record |
| Exploitation | CISA ADP KEV block where listed; SSVC Exploitation decision point | KEV-listed CVEs prioritised | exploitedSince; exploited dashboard |
| Likelihood | Not in schema | Not its role | epss |
| Dates | datePublic (in containers), dateUpdated (cveMetadata) | NVD publish and modified dates | datePublished, dateUpdated |
| Cross-reference | references | references | aliases, references, assigner |
Each source in depth
CVE Programme records
- What it protects against: naming confusion. One ID per vulnerability lets vendors, scanners and defenders refer to the same flaw.
- Architecture in words: a federated model. CNAs assign IDs within their scope and write the record; Roots oversee CNAs in a domain or region; ADPs such as CISA add containers beside the CNA's data. Records flow out through the CVE Services API and the public cvelistV5 repository.
- Failure modes: quality varies by CNA. Some records carry precise version ranges, CVSS and CWE; others have only a description and a product name. A REJECTED record left open in a scanner produces findings that can never be closed.
- Controls: parse the full 5.x schema, not the retired 4.0 feeds; act on state changes; store the CNA's shortName from providerMetadata so you can measure record quality per CNA. The Becoming a CVE Numbering Authority Guide covers the CNA side.
CISA Vulnrichment (ADP layer)
- What it adds: three SSVC decision points for every CVE it analyses (Exploitation, Automatable, Technical Impact), CWE and CVSS for higher-risk CVEs when evidence allows, and a KEV block for catalogued CVEs. It stopped adding CPE strings on 10 December 2024.
- Precedence rule built in: CISA never overwrites the CNA container, removes its own CWE, CVSS or CPE if the CNA later adds them, and treats both being present as an ADP error where CNA data wins.
- Control: check the SSVC version string before parsing, because a change to the decision tree changes the version.
NVD
- What it protects against: thin CNA records. NVD analysis historically supplied the CPE applicability statements many scanners matched on.
- Current behaviour: from 15 April 2026 NIST enriches CVEs in the CISA KEV catalogue (aiming for one business day from receipt), CVEs in software used within the federal government and CVEs in critical software as defined under Executive Order 14028. Others remain in the NVD labelled "Lowest Priority - not scheduled for immediate enrichment", and users may request enrichment by email. Backlogged CVEs with an NVD publish date before 1 March 2026 moved to "Not Scheduled", except KEV entries.
- Failure modes: a scanner that only matches on NVD CPE data will not see unenriched CVEs; a dashboard that sorts on "NVD score" will show blanks for CVEs where the CNA score is now the only score.
- Controls: treat "no NVD enrichment" as a recorded state, not an error; fall back to the CNA affected array and vendor advisory for matching.
EUVD
- What it protects against: dependence on one national source, and the gap for vulnerabilities coordinated by European CSIRTs.
- Architecture in words: NIS2 Article 12(2) requires the database to include information describing the vulnerability, the affected ICT products or services and its severity, and the availability of patches or, where none exist, guidance from competent authorities or CSIRTs. ENISA states that CVE data, vendor advisories and the CISA KEV catalogue are transferred automatically, and the platform uses the open-source Vulnerability-Lookup software to correlate sources. It offers critical, exploited and EU coordinated dashboards.
- Failure modes: treating the EUVD identifier as a new vulnerability creates duplicates; treating its score as authoritative when the CNA has since revised the record creates drift.
- Controls: join on aliases to the CVE ID; compare dateUpdated with the CVE record before accepting a changed value.
Source-of-truth policy: field precedence
Write this as a policy and implement it in the ingestion layer. The order below reflects who is closest to the product and how each source describes its own precedence.
| Field | First choice | Second | Third | Notes |
|---|---|---|---|---|
| Existence and status | CVE record state | Vendor advisory | EUVD alias | REJECTED closes findings everywhere |
| Affected versions | Vendor advisory | CNA affected array | NVD CPE applicability | For open source, the project's own advisory or package ecosystem data |
| Fixed version and mitigation | Vendor advisory | CNA references | EUVD patch or CSIRT guidance | Record the advisory URL as evidence |
| Severity (base) | CNA metrics | CISA ADP CVSS | NVD score, then EUVD baseScore | Store all with source and version; never average |
| Weakness (CWE) | CNA problemTypes | CISA ADP CWE | NVD CWE | Used for trend analysis, not prioritisation |
| Exploitation | CISA KEV catalogue | EUVD exploitedSince | CISA SSVC Exploitation, vendor statement | Any credible source marks exploited; do not wait for all three |
| Likelihood | EPSS | EUVD epss field | None | See the EPSS Guide |
Reconciling conflicting records
Conflicts are normal: sources update at different times and score different CVSS versions. A repeatable workflow keeps them from becoming arguments.
- Normalise the key. Join every record on the CVE ID; map EUVD IDs through aliases. Keep the EUVD ID as an attribute for EU reporting.
- Compare versions of the same metric only. A CVSS v4.0 score from the CNA and a v3.1 score from another source are not in conflict; store both with baseScoreVersion or the metric key.
- Check freshness. Compare dateUpdated across sources. If the newer value comes from a lower-precedence source, raise a review rather than overwriting.
- Apply precedence. Use the table above to choose the effective value, and keep the losing values with their source for audit.
- Escalate material conflicts. Where the effective severity crosses your SLA threshold in one source but not another, or affected versions disagree on whether your installed version is vulnerable, route to an analyst and, if needed, ask the vendor or CNA.
- Raise exploitation conservatively. If any trusted source marks the CVE exploited, treat it as exploited until disproved.
- Record the decision. Store the effective value, the source, the time and the analyst, so the SLA clock and the evidence agree.
Worked scenario: a library CVE is published with a CNA CVSS v4.0 score of 6.9 and affected versions up to 2.3.1. The NVD labels it Lowest Priority and adds nothing. A week later the EUVD record shows a CVSS 3.1 score of 8.1 and an exploitedSince date, and the CVE appears in the CISA KEV catalogue. Under the policy, exploitation is now true (KEV), the base severity stays the CNA's v4.0 score with the 3.1 score stored alongside, the NVD will now prioritise enrichment because of the KEV listing, and the SLA moves to your exploited-vulnerability deadline from the KEV date. The vendor advisory confirms 2.3.2 as the fix and becomes the evidence link. See the CISA KEV Catalogue Guide for KEV deadlines.
Feeding scanners and platforms
| Approach | Strength | Risk | Suits |
|---|---|---|---|
| Scanner vendor feed only | Least effort | Opaque source choice; gaps follow the vendor's dependence on NVD | Small teams, with a written question to the vendor about sources |
| CVE records plus vendor advisories, NVD as optional enrichment | Fastest, closest to the product | Needs schema parsing and version matching without CPE | Teams with tooling engineers |
| Multi-source with precedence (CVE, CISA ADP, NVD, EUVD, KEV, EPSS) | Resilient to any one source degrading | Build and maintenance effort; conflict handling | Large enterprises, MSSPs, product security teams |
- Ask your scanner vendor: which sources it uses for matching, how it handles CVEs without NVD CPE data, which CVSS version it displays and whether it reads ADP containers.
- Small organisations: rely on the scanner, but subscribe to vendor advisories for critical products and to KEV, and check EUVD exploited entries if you are a NIS2 entity.
- Large organisations: build the precedence layer once in the vulnerability platform and let scanners supply detection only. Container and package findings carry their own matching problems, covered in the Backported Security Fixes Guide.
Governance and funding risk
Each layer has a different single point of failure. The CVE Programme depends on a US government sponsor and contract, which came close to lapsing in April 2025; the NVD depends on NIST resourcing, which led to the April 2026 prioritisation; the EUVD depends on upstream data and ENISA operations, but has an EU legal basis in NIS2 Article 12. ENISA's move to CVE Root for European entities in November 2025, with 20 CNAs under it according to ENISA, spreads assignment capacity across regions. The practical response is the same for all three: keep raw copies of records you rely on, keep provenance per field, and test that your pipeline still works when any one source stops updating.
Detection, measurement and framework mapping
- Share of open findings whose effective severity came from each source (watch for sudden shifts).
- Median hours from CVE publication to first effective severity in your platform.
- Open findings with no NVD enrichment that were still matched to assets (proves the fallback works).
- Material conflicts raised and median days to resolve.
- Findings left open against REJECTED CVEs (target zero).
| Activity | NIST CSF 2.0 | NIST SP 800-53 Rev. 5 |
|---|---|---|
| Identify and record vulnerabilities from multiple sources | ID.RA-01 | RA-5 Vulnerability Monitoring and Scanning |
| Receive and act on advisories and alerts | ID.RA-08 | SI-5 Security Alerts, Advisories, and Directives |
CIS Control 7 covers the wider programme; see the CIS Control 7 Continuous Vulnerability Management Guide.
Common failure patterns
- Last-write-wins ingestion: the effective score flips each time a source updates, and SLA clocks restart.
- NVD-only matching: CVEs labelled Lowest Priority never match an asset, so real exposure goes unreported.
- Averaging or mixing CVSS versions: a v3.1 and a v4.0 score compared as if on one scale.
- Duplicate findings from EUVD IDs: the same flaw counted twice because aliases were not joined.
- Ignoring ADP containers: CISA's SSVC and KEV data sit in the record but the parser reads only the CNA container.
- Waiting for consensus on exploitation: holding a patch until every source agrees, when KEV or the EUVD exploited view already says yes.
Practitioner checklist
- Write a source-of-truth policy with field precedence and get it approved.
- Parse CVE Record Format 5.x, including adp containers and the REJECTED state.
- Store every score with its source, CVSS version and timestamp.
- Match assets with CNA affected data and vendor advisories when NVD enrichment is absent.
- Join EUVD records through aliases and keep the EUVD ID for EU reporting.
- Mark exploitation from KEV and the EUVD exploited view, whichever comes first.
- Ask scanner vendors in writing which sources they use and how they handle unenriched CVEs.
- Track the five metrics above monthly and review the policy each year.
Standards and references
- CVE Program: CVE Record Format JSON schema
- NIST: National Vulnerability Database programme page and status updates
- CISA: Vulnrichment
- EUR-Lex: Directive (EU) 2022/2555 (NIS2), Article 12
- ENISA: Consult the European Vulnerability Database
- ENISA: ENISA scales up its role in the CVE Program
- ENISA: EUVD API, exploited vulnerabilities endpoint
Related NHI Mgmt Group resources: CVE Programme and NVD Backlog Guide · EU Vulnerability Database (EUVD) · CISA KEV Catalogue Guide · CVSS v4 Guide · Risk-Based Vulnerability Management Guide