If you ship software or hardware and you are not a CVE Numbering Authority (CNA), someone else decides when your vulnerabilities get CVE IDs and what the CVE Record says about them. A CNA is an organisation authorised by the CVE Program to assign CVE IDs and publish CVE Records within an agreed scope. Becoming one costs no fee and needs no contract, but the CNA Operational Rules (version 4.2.6, approved by the CVE Board on 14 October 2026) carry real obligations: a published scope, response times, record content, publication deadlines and account management. This guide is for PSIRT leads, open source maintainers and product security managers deciding whether to become a CNA and how to run one well. It covers the decision, the onboarding steps, the rules that matter day to day, record quality, tooling and credentials, and how to measure the programme. How the wider CVE Programme and NVD enrichment work is covered in the CVE Programme and NVD Backlog Guide; this page is about the CNA role itself.
Key takeaways
- The entry bar is low, the operating bar is not. The three requirements are a public vulnerability disclosure policy, a public source for new vulnerability disclosures and agreement to the CVE Terms of Use; the rules then add MUST-level duties on scope, contacts, timing and content.
- Being a CNA gives you first refusal. The CNA with the most appropriate scope must be given the first opportunity to assign; if it says no or stays silent for 72 hours on a publicly disclosed issue, a Root can direct another CNA to assign.
- The clock starts at public disclosure. Records should be published within 24 hours and must be published within 72 hours of the CNA publicly disclosing an ID it assigned, or a CNA of Last Resort may publish instead.
- Record quality is now your problem. A record must name at least one affected product, a problem type (CWE recommended), an English description and a public reference that is not the record itself, and downstream tools read your CNA container directly.
- CVE Services accounts are privileged machine credentials. API keys that reserve IDs and publish records should be vaulted, rotated and deactivated promptly, which rule 3.2.1.2 requires for inactive users.
Should your organisation become a CNA?
The CVE Program lists the benefits as demonstrating mature vulnerability management, controlling the publication process for vulnerabilities in your scope, assigning IDs without sharing embargoed details with another CNA and streamlining disclosure. The cost is staff time and the discipline the rules demand.
| Situation | Become a CNA? | Reasoning |
|---|---|---|
| Supplier publishing several security advisories a year, with a PSIRT | Yes | You already do the work; becoming a CNA removes the dependency on another CNA and lets you publish the record with your advisory. |
| Widely used open source project with a security team | Usually yes, or via a foundation CNA | A foundation or hosting platform may already be a CNA whose scope covers you; check the List of Partners before applying. |
| Hosted service or SaaS only | Consider it | The rules allow records for hosted services and require the exclusively-hosted-service tag when every affected product exists only as a hosted service. |
| Supplier with one or two vulnerabilities in several years | Probably not yet | Request IDs from your Root's CNA of Last Resort or a coordinator; revisit when volume grows or regulation (for example the EU Cyber Resilience Act) raises the stakes. |
| Research firm or bug bounty platform | Yes, with a research scope | Rule 3.1.8: define scope as vulnerabilities you research or coordinate, unless another CNA has more specific scope. |
| No public disclosure policy or advisory page | Not until fixed | Both are entry requirements; build them first. See the Bug Bounty and Vulnerability Disclosure Guide. |
Small versus large: a small supplier can run a CNA with two named people, a shared mailbox and a web client; the rules ask for at least two individuals or addresses for the administrative contact (SHOULD, rule 3.2.2.4). A large supplier with many brands needs a scope that survives acquisitions (rule 3.1.6 suggests wording such as "All XYZ (formerly A and B) Products") and may need several CNA identities, each with its own scope definition (rule 3.1.12.1).
Roles in the CVE Program hierarchy
| Role | What it does | Relevance to a new CNA |
|---|---|---|
| CVE Board | Oversees operations and approves the CNA Operational Rules and exceptions. | Only the Board can grant an exception to the rules (rule 2.3.2) or revoke CNA status. |
| Secretariat (currently MITRE) | Administrative support and programme infrastructure; final review of scope definitions. | Receives your registration through the CVE Program Web Forms. |
| Top-Level Root (CISA, MITRE) | Reports to the Board and manages a hierarchy of Roots and CNAs. | Approves Root additions to the rules within its hierarchy. |
| Root (for example ENISA, Google, INCIBE, JPCERT/CC, Red Hat, Thales Group, CERT@VDE, CISA ICS) | Manages CNAs in a technical or geographic area. | Approves your scope, reviews rule infractions and resolves first-refusal failures. |
| CNA of Last Resort (CNA-LR) | Broad-scope CNA that assigns when no other CNA does. | May publish a record for your ID if you miss the 72-hour deadline. |
| Authorized Data Publisher (ADP) | Enriches records in a separate ADP container; cannot modify the CNA container. | The CISA ADP adds SSVC decision points, KEV data and, for some records, missing CVSS, CWE or CPE. |
The CVE Program structure page describes these roles as current at the time of checking (27 October 2026). The list of Roots changes, so confirm yours on the List of Partners.
Onboarding: the five steps
- Contact: use the CVE Program Web Forms and choose "Register to Become a CNA".
- Register: the form collects organisation name, country, sector, representative contacts, the estimated annual number of CVE IDs, the expected scope, disclosure policy and advisory links, and availability for a one-hour orientation meeting.
- Meet requirements: a public vulnerability disclosure policy, a public source for new disclosures and agreement to the CVE Terms of Use. Your scope definition must be approved by your Root or TL-Root, with final review by the Secretariat, before you are authorised (rules 3.1.1.1 and 3.1.1.2).
- Onboard: watch the CVE lightning talk, attend an introductory session and successfully complete practice examples.
- Join: you are added to the List of Partners, with your scope, public contact, policy URL and advisory distribution point, and announced as a new partner.
Preparation that shortens onboarding
- Write the scope first: rule 3.1.3 recommends broad supplier wording ("Supplier X's products only") and rule 3.1.4 forbids a scope that only lists what is not covered. Decide now whether you cover end-of-life products (rule 3.1.11.1).
- Publish the assignment guidance: rule 3.2.6.1 requires guidance on how you assign CVE IDs and publish records within your disclosure process, and rule 3.2.6.2 requires a URL to it. ISO/IEC 29147 is a good model for the disclosure side; see the ISO/IEC 29147 and 30111 Vulnerability Disclosure and Handling Guide.
- Remove restrictive terms: rule 3.2.4.2 says the method to request an ID must not be encumbered by terms that restrict assignment or publication; a bug bounty NDA that blocks this effectively waives your first refusal.
- Stand up the advisory page: supplier CNAs must have at least one public distribution point that references CVE IDs, should not require login and must not impose restrictive terms of use (rules 4.5.2.2 and 4.5.2.4). Machine-readable advisories help; see the CSAF Security Advisories Guide.
Day-to-day obligations under the CNA Operational Rules
The rules use MUST, SHOULD and MAY with their usual meaning. The table lists the obligations that most often decide whether a CNA runs smoothly. Rule numbers are from version 4.2.6.
| Area | Rule | Requirement |
|---|---|---|
| Administrative contact | 3.2.2.1, 3.2.2.2, 3.2.2.5 | MUST provide an administrative POC with email addresses and phone numbers, and MUST respond in a timely manner; automated replies do not count. |
| Public contact | 3.2.3.1, 3.2.4.1 | MUST publish a public POC in the List of Partners and respond in a timely manner to assignment requests within scope. |
| Accounts | 3.2.1.2 | MUST disable, deactivate or otherwise manage inactive user accounts in a timely manner. |
| Programme communications | 3.2.4.5 | MUST follow official communications, currently the CVE CNA Discussion mailing list. |
| First refusal | 4.2.1.2 | For publicly disclosed issues, declining or not responding within 72 hours hands the decision to a Root, which must direct assignment within 72 hours. |
| Upstream dependencies | 4.2.16.6 | A supplier CNA that finds a vulnerability in an upstream dependency MUST notify and defer first refusal to the upstream supplier if it is a CNA. |
| Separate vulnerabilities | 4.2.6, 4.2.11 | MUST assign different IDs to separate, independently fixable vulnerabilities (changed from SHOULD to MUST in 4.2.6). |
| Shared code | 4.2.13 | MUST NOT assign more than one ID when several products are vulnerable because they share the same vulnerable code. |
| Disclosure | 4.2.4, 4.2.5 | An assigned ID must be publicly disclosed or rejected; do not assign IDs for issues you never intend to disclose. |
| Publication timing | 4.5.1.3, 4.5.1.4, 4.5.1.6 | SHOULD publish within 24 hours and MUST publish within 72 hours of disclosing an ID you assigned; SHOULD publish within 72 hours of learning another party disclosed it. |
| Unused IDs | 4.5.3.5 | MUST reject unused or unpublished IDs. |
| Reporter | 4.3.3 | MUST provide the CVE ID to the person who reported the vulnerability. |
| End of life | 4.2.17.1 | SHOULD assign for EOL products; MUST tag records covering only products that no longer receive fixes as unsupported-when-assigned. |
Vulnerability determination calls you will face
- Insecure defaults: rule 4.1.3 says CNAs should treat insecure default configuration as a vulnerability, particularly defaults that enable direct exploitation or surprising behaviour.
- Dependency updates: updating a library is never a new vulnerability (rule 4.1.11); your advisory should reference the library's CVE ID.
- Technology type: cloud, AI or machine learning cannot be the sole basis for deciding whether to assign (rule 4.2.3).
- Unclear cases: after a good faith effort, err on the side of assignment (rule 4.4.3), and consult peer CNAs or your Root on unusual judgments (rule 4.4.2).
- Not vulnerabilities: brute-force denial of service, general purpose malicious code and deliberately vulnerable training software such as OWASP WebGoat (rules 4.1.5, 4.1.7, 4.2.18).
Writing CVE Records that downstream tools can use
CVE Services uses the CVE Record Format, current version 5.2.0 at the time of checking. In the schema, a published CNA container requires providerMetadata, descriptions, affected and references; optional fields include problemTypes, metrics, cpeApplicability, solutions, workarounds, exploits, timeline, credits and tags. The rules then add content requirements beyond the schema.
| Element | Rule | Practical standard |
|---|---|---|
| Affected products | 5.1.3, 5.1.4, 5.1.8 | At least one product marked affected or unknown, never only unaffected. Use version ranges with versionType and a non-inclusive limit, and list fixed versions (5.1.5). |
| Problem type | 5.1.7, 5.1.7.1 | State the cause using CWE; "not enough information" is allowed only temporarily. See the CVE, CWE and CPE Guide. |
| Description | 5.2.1 to 5.2.6 | English description that summarises the vulnerability, does not contradict the structured data and does not credit people or tools (use the credits element instead). |
| References | 5.1.10, 5.3.1, 5.3.3 | At least one freely available public reference with vulnerability specific information that exists before or with the record; the record must never be the first disclosure. |
| Severity | Optional metrics element | Supply a CVSS vector yourself rather than leaving it to enrichment. See the CVSS v4 Guide. |
| Tags | 5.1.11 | exclusively-hosted-service and unsupported-when-assigned where they apply; "disputed" uses the programme's formats (4.5.3.10). |
| Test data | 5.4.1, 5.4.2 | Use the CVE-1900- prefix for examples and testing; CVE Services does not accept it. |
Why quality matters more now: as the companion guide explains, the NVD enrichment backlog means many consumers read the CNA container first. A record with precise affected ranges, a CWE and a CVSS vector is immediately usable for scanning and prioritisation, and the CISA ADP will add SSVC decision points alongside it. Link fixes and advisories that stay available: rule 4.5.2.6 asks CNAs to consider redirects or archiving for long-term URL availability.
Changing a published record
- Update: when a fix becomes available or material facts change (4.5.3.1), and to correct materially wrong information (4.5.3.2).
- Reject: with an explanation, for duplicates, wrong IDs or no vulnerability (4.5.3.6, 4.5.3.7); merging assignments requires rejecting the redundant IDs (4.5.3.8).
- Disputes: follow the CVE Program Policy and Procedure for Disputing a CVE Record, after first trying to resolve the disagreement directly (4.6.2, 4.6.3).
Tooling and the credentials behind it
CVE Services is the programme's self-service interface for reserving IDs and publishing records. The CVE website lists web and command line clients, including Vulnogram and cvelib, and documents API calls that authenticate with three headers: CVE-API-USER, CVE-API-ORG and CVE-API-KEY. Those keys can reserve IDs and publish public records in your organisation's name, which makes them high-impact machine credentials.
- One user per person or pipeline: do not share a single key across the PSIRT. Named users make audit possible and let you remove one person without breaking everyone.
- Vault the keys: store CVE Services API keys in a secrets manager, inject them into the publishing pipeline at runtime and never commit them to the advisory repository. See the Secrets Management Guide.
- Offboard promptly: rule 3.2.1.2 requires timely management of inactive accounts; tie CVE Services user removal to your HR leaver process.
- Separate test from production: the programme must provide a non-production environment for format changes (rule 4.5.5.2); point automation at it before a schema version change.
- Protect embargoed data: reserved IDs and draft records describe unfixed vulnerabilities. Limit who can read the PSIRT tracker and the draft JSON as tightly as the CVE Services account itself.
Running the programme: a phased rollout
- Weeks 1 to 4, foundations: publish the disclosure policy, the assignment guidance URL and the advisory distribution point; agree the scope wording internally; name two administrative contacts with phone numbers.
- Weeks 4 to 8, registration and onboarding: submit the web form, attend orientation, complete the practice examples and get scope approval from your Root.
- Pilot: handle your first few assignments with the Root available for questions; publish records in the same change as the advisory so the 24-hour target is met by design.
- Automate: generate CVE Record JSON from the PSIRT tracker, validate it against the schema in CI and publish through CVE Services with a vaulted key.
- Steady state: quarterly review of reserved but unpublished IDs (reject the unused ones), scope changes after acquisitions and record quality metrics.
Rollback and contingency: if the PSIRT cannot meet its obligations for a period, tell your Root early. The rules provide fallbacks (a CNA-LR can assign or publish, and record ownership can be transferred), but serious or repeated failures are reviewed by the Root and the CVE Board may revoke CNA status (rule 1.2.3.1).
Detection and measurement
- Median hours from public advisory to published CVE Record (target under 24; never over 72).
- Percentage of assignment requests answered within your documented response time.
- Count of reserved IDs older than 12 months without publication or rejection (target zero).
- Percentage of records with affected version ranges, fixed versions, a CWE and a CVSS vector.
- Number of records later updated for materially incorrect data, and disputes received.
- CVE Services users without activity in 90 days (target zero after review).
Framework mapping
| Framework | Reference | How the CNA role supports it |
|---|---|---|
| NIST CSF 2.0 | ID.RA-08 | Processes for receiving, analysing and responding to vulnerability disclosures. |
| NIST SP 800-53 Rev. 5 | RA-5(11) Public Disclosure Program | Public reporting channel and handling of reported vulnerabilities. |
| NIST SP 800-53 Rev. 5 | SI-2 Flaw Remediation; IA-5 Authenticator Management; AC-2 Account Management | Fix tracking linked to records; management of CVE Services keys and users. |
| ISO/IEC 27001:2022 | Annex A 8.8 Management of technical vulnerabilities | Supplier-side handling and publication of vulnerability information. |
| ISO/IEC 29147 and 30111 | Disclosure and handling processes | The disclosure policy and handling process the CNA requirements assume. |
For product suppliers in the EU, the CRA's vulnerability handling and reporting duties sit alongside, not inside, the CVE rules; see the EU Cyber Resilience Act Vulnerability and Incident Reporting Guide.
Common failure patterns
- Advisory published, record forgotten: the advisory goes out and the record sits in RESERVED state past 72 hours, so a CNA-LR may publish it and ownership may move.
- Scope written for today's product list: an acquisition brings in products outside the scope, another CNA assigns for them and you lose control of the record text.
- One CVE for a bundle of bugs: combining independently fixable issues into one ID breaks rule 4.2.11 and makes scanners report the wrong fixed versions.
- Thin records: no version ranges, no CWE, a description copied from the advisory title; consumers cannot act until someone else enriches it.
- Shared API key in a script: a single CVE Services key in a repository or a former employee's laptop that can still publish records in your name.
- Reporter left in the dark: not sending the reporter the CVE ID (a MUST under 4.3.3) damages trust with the research community you rely on. See the PSIRT Guide for reporter handling.
Practitioner checklist
- Check the List of Partners for an existing CNA whose scope already covers your products.
- Publish the disclosure policy, CVE assignment guidance URL and a login-free advisory page.
- Draft a broad scope definition that covers future products and states your end-of-life approach.
- Name at least two administrative contacts with email and phone, and a monitored public contact.
- Register through the CVE Program Web Forms and complete orientation and practice examples.
- Build record generation and schema validation into the advisory pipeline so records ship with advisories.
- Issue named CVE Services users, vault the API keys and tie removal to the leaver process.
- Review reserved IDs quarterly and reject the unused ones.
- Subscribe the team to the CVE CNA Discussion mailing list and track rule version changes.
Standards and references
- CVE Program: CNA Operational Rules
- CVE Program: Become a Partner and onboarding process
- CVE Program: Program Structure
- CVE Program: CVE Services
- CVE Program: CVE Record Format schema
- CVE Program: Authorized Data Publishers
- CVE Program: Frequently Asked Questions
- NIST: Cybersecurity Framework 2.0
Related NHI Mgmt Group resources: CVE Programme and NVD Backlog Guide · PSIRT Guide · How to Read a CVE Advisory Guide · VEX Guide