An SBOM tells a customer which components are inside a product, and a scanner then reports every CVE in those components, most of which do not affect the product at all. Vulnerability Exploitability eXchange (VEX) is the machine-readable way for a supplier to say, per product and per vulnerability, whether the product is affected, not affected, fixed or still under investigation, and why. Done well, it removes large volumes of false positives from customers' queues; done badly, it hides real risk behind an unverified "not affected". This guide covers the VEX status model, the three main formats (CSAF VEX, OpenVEX and CycloneDX), how producers should write defensible justifications, and how consumers should ingest, verify and act on VEX. It is written for product security teams, PSIRTs, vulnerability management leads and engineers wiring VEX into scanners.
Key takeaways
- VEX is a statement, not a scan result. Each statement joins a product, a vulnerability and a status at a point in time, made by a named author. Its value depends entirely on how trustworthy that author and analysis are.
- Four statuses carry across formats. Not affected, affected, fixed and under investigation are the core states in OpenVEX and the CSAF VEX profile, and CycloneDX expresses the same ideas through its analysis states.
- "Not affected" must say why. OpenVEX requires a justification label or an impact statement for every not affected statement, and CSAF uses the same five justification flags. Treat a not affected claim without a reason as invalid.
- VEX is a stream that changes. Statements are meant to evolve from under investigation to affected to fixed. Consumers must take the latest statement per product and vulnerability, not the first one they saw.
- Consumers should suppress, not delete. Use VEX to filter scanner output, but keep the suppressed finding, its source document and its justification so you can audit and reverse the decision.
What VEX is and where it fits
VEX is a companion to the software bill of materials. The SBOM lists components; vulnerability databases map components to CVEs; VEX adds the supplier's assessment of whether each CVE is exploitable in the product as built and shipped. OpenVEX describes itself as meeting the "Minimum Requirements for VEX" document published in April 2023 by the VEX working group coordinated by CISA, which is the common baseline across formats. See the SBOM Guide for SBOM generation and exchange, and the Reachability Analysis Guide for the analysis techniques that often sit behind a VEX statement.
The OpenVEX specification frames a valid statement as having four data points, like parts of a sentence:
- Product: one or more products, usually identified with package URLs (purl) or other identifiers that can be correlated with an SBOM.
- Vulnerability: an identifier, typically a CVE ID.
- Status: the impact of the vulnerability on the product.
- Timestamp: when the statement was made, because statements are only meaningful relative to others on the same subject.
CISA's guidance on when to issue VEX notes that whether, and when, to publish VEX is a business decision for most suppliers. In practice it is most valuable for widely deployed products, for high-profile CVEs that trigger mass customer enquiries, and for regulated customers who must evidence how they handled each SBOM finding.
The status model
| Meaning | OpenVEX status | CSAF product_status | CycloneDX analysis.state | Required extra information |
|---|---|---|---|---|
| Not affected: no remediation needed | not_affected | known_not_affected | not_affected (or false_positive where the match itself was wrong) | Justification label or impact statement |
| Affected: action recommended | affected | known_affected | exploitable | Action statement (OpenVEX) or remediation (CSAF) |
| Fixed in these versions | fixed | fixed (also first_fixed) | resolved or resolved_with_pedigree | Version or product identifiers that contain the fix |
| Not yet known | under_investigation | under_investigation | in_triage | Follow-up statement when analysis completes |
CSAF also defines first_affected, last_affected and recommended product statuses, which are useful in advisories but outside the four core VEX states. CycloneDX's resolved_with_pedigree means the fix is evidenced in the component's pedigree with verifiable commit history or diffs, which is stronger evidence than a plain claim.
Justifications for "not affected"
The justification is the part of VEX that most needs scrutiny. OpenVEX and the CSAF flag labels share the same five values; CycloneDX uses its own, more granular list.
| OpenVEX and CSAF label | Meaning | Evidence a consumer should expect |
|---|---|---|
component_not_present | The vulnerable component is not included in the product | SBOM showing the component absent; often caused by a scanner mis-match |
vulnerable_code_not_present | The component is included but the vulnerable code is not, for example excluded at build time | Build configuration or patch reference |
vulnerable_code_not_in_execute_path | The vulnerable code is present but cannot be executed as the product uses it | Reachability analysis, call graph or configuration explanation |
vulnerable_code_cannot_be_controlled_by_adversary | The vulnerable code runs but an attacker cannot control the input | Data flow reasoning; the OpenVEX spec warns this can be difficult to prove conclusively |
inline_mitigations_already_exist | Built-in protections prevent exploitation and cannot be disabled or subverted | Description of the mitigation and why it cannot be configured off |
CycloneDX justifications are code_not_present, code_not_reachable, requires_configuration, requires_dependency, requires_environment, protected_by_compiler, protected_at_runtime, protected_at_perimeter and protected_by_mitigating_control. Note that the last three, and requires_configuration and requires_environment, depend on how the customer deploys the product. A "not affected" that is only true if the customer has a web application firewall or a default setting is really a conditional statement, and the consumer must check whether the condition holds in their environment.
The OpenVEX specification includes a well-known example from Log4Shell: Spring Boot stated that its logging starter was not affected because users are only affected if they switch the default logging system to Log4J2, using the vulnerable_code_not_in_execute_path justification with an impact statement. It shows both the value of VEX (millions of scanner alerts answered at once) and its limits (customers who had changed the default were affected).
The three main formats
| Aspect | CSAF 2.0 VEX profile | OpenVEX | CycloneDX VEX |
|---|---|---|---|
| Steward and status | OASIS Standard (18 November 2022), Errata 01 approved 26 January 2024 | OpenVEX specification v0.2.0 (community project, checked 27 October 2026) | CycloneDX specification (schema 1.6 reviewed for this guide) |
| Shape | Full advisory document with document.category csaf_vex, product tree, vulnerabilities, product status, flags, threats and remediations | Lightweight JSON-LD document with metadata and a list of statements | vulnerabilities array with analysis (state, justification, response, detail) and affects; standalone or inside a BOM |
| Required for not affected | Impact statement through a flag or a threat of category impact | justification or impact_statement | analysis.state, ideally with justification and detail |
| Required for affected | Remediation (categories mitigation, no_fix_planned, none_available, vendor_fix, workaround) | action_statement | response values such as update, workaround_available, will_not_fix |
| Distribution | CSAF provider metadata (provider-metadata.json) discoverable at a well-known URL, plus aggregators; TLP labels in the distribution field | Files, OCI attestations, embedding in other formats (in-toto, CSAF, CycloneDX) | With or alongside the SBOM; BOM exchange APIs |
| Best fit | Vendors that already publish CSAF advisories and customers with CSAF tooling | Container images, open source projects and CI pipelines | Organisations standardised on CycloneDX SBOMs |
SPDX 3.0 also models VEX in its Security profile through VexAffectedVulnAssessmentRelationship, VexFixedVulnAssessmentRelationship, VexNotAffectedVulnAssessmentRelationship and VexUnderInvestigationVulnAssessmentRelationship. For CSAF advisory structure in general, see the CSAF Guide.
Decision guidance: pick the format your customers and your own SBOM tooling already consume. A large vendor with an established PSIRT and CSAF feed should publish CSAF VEX. A team shipping container images through a registry will find OpenVEX attestations simpler. A software producer whose customers ask for CycloneDX SBOMs can carry VEX in the same ecosystem. Consumers with mixed suppliers need tooling that reads all three; several open source scanners already do, as described below.
Producing VEX: a workflow for suppliers
- Start with accurate SBOMs. VEX statements must reference product identifiers that customers can correlate with the SBOM. Mismatched purls or versions make VEX useless.
- Triage each new CVE against each product version. Publish
under_investigationquickly for high-profile CVEs so customers know analysis is under way. - Record the evidence behind every not affected claim. Store the reachability result, build flag or code reference internally, even if the public statement only carries the label and a short impact statement.
- Prefer narrow, version-specific statements. State exactly which product versions and architectures are covered. Avoid blanket statements for "all versions".
- Update, never silently edit. Increment the document version (OpenVEX requires this when any content changes), add new statements as status changes, and keep the history.
- Sign and publish where customers will find it. Use signed attestations or a CSAF trusted provider location so consumers can verify the author. The OpenVEX tool
vexctlsupportscreate,merge, andattest --attach --signfor container images. - Review and test. Peer-review not affected statements for high and critical CVEs, and validate documents against the format schema before publishing; CSAF defines mandatory tests, including one for VEX product status.
Rollback: if a not affected statement turns out to be wrong, publish a new statement with status affected and an action statement, notify customers through the advisory channel, and record the correction. Never remove the old statement without a newer one replacing it.
Consuming VEX: a workflow for customers
- Collect VEX alongside SBOMs. Require VEX or an equivalent per-CVE status from critical suppliers in contracts, and store documents with their source and signature.
- Verify the author. Accept VEX only from the supplier of the product or a party you trust, verified by signature or trusted distribution location.
- Apply statements in your scanner. Grype supports OpenVEX for filtering and augmenting results; Trivy (VEX support marked experimental) accepts
--vexwith a VEX repository, a local file, OCI attestations or SBOM references, with the order setting priority;vexctl filterapplies VEX to SARIF scan results. - Suppress with a record. Mark matched findings as suppressed with the VEX document ID, statement timestamp and justification, so auditors and analysts can see why the finding was closed.
- Check conditional justifications. For configuration, environment or perimeter based justifications, confirm the condition holds in your deployment before accepting.
- Re-evaluate when statements change. Process new VEX versions and re-open findings whose status moves to affected.
- Escalate KEV-listed CVEs. For vulnerabilities in the CISA KEV Catalogue Guide scope, require stronger evidence before accepting not affected, and consider your own verification.
Internal VEX for your own applications
VEX is not only for suppliers and customers. Organisations that build software can use the same model internally to record triage decisions on their own SBOMs, so that the same CVE is not re-triaged by every team, every sprint. A central product security team publishes internal VEX statements for shared base images and libraries, and application teams inherit them.
- Single source of triage: store internal VEX documents in a repository next to the SBOMs they describe, with pull request review for every not affected statement.
- Expiry: give internal not affected decisions a review date (for example 90 days, or on the next major release), because code paths and configuration change.
- Scope: tie statements to image digests or exact versions, so a rebuild with a new dependency version does not inherit an outdated decision.
- Audit evidence: internal VEX gives auditors a clear record of why open scanner findings were accepted, which is often requested under vulnerability management policies. See the Vulnerability Management Guide.
Where identity fits
A VEX document is only as trustworthy as the identity that signed it. OpenVEX requires an author that is an individual or organisation, and recommends that author identity is cryptographically verifiable. Signing keys or workload identities used in CI to produce attestations are non-human identities that need ownership, rotation and least privilege: a stolen signing identity could publish "not affected" for an exploited vulnerability across every customer. Treat VEX signing like release signing. See the Software Supply Chain Security Guide.
Detection and measurement
- VEX coverage: share of SBOM findings from critical suppliers that have a VEX status, by supplier.
- Suppression volume: number and percentage of scanner findings suppressed by VEX, with a breakdown by justification label.
- Reversal rate: how often a not affected statement is later changed to affected. A non-zero rate is normal; a high rate for one supplier is a trust signal.
- Time to status: for producers, hours from CVE publication to first VEX statement (under investigation) and to final status.
- Unverified statements: count of VEX documents accepted without signature or trusted-source verification. The target is zero.
Monitor for unexpected changes: a supplier's VEX feed suddenly marking many critical CVEs not affected, or statements signed by an unfamiliar key, should be investigated as a possible supply chain compromise (T1195.002).
Common failure patterns
- Not affected without evidence. Statements carry a justification label but nobody can produce the analysis behind it. Keep the evidence internally, and as a consumer ask for it on critical CVEs.
- Product identifiers that do not match the SBOM. The VEX names a product or version string that the scanner cannot correlate, so nothing is suppressed, or worse, the wrong product is.
- Stale statements. Consumers apply an old not affected after the supplier has published a newer affected statement because they do not process updates.
- Deleting findings instead of suppressing them. Removing findings breaks audit trails and makes reversal impossible.
- Accepting conditional statements blindly. A statement that depends on a default configuration is accepted by a customer who changed that configuration, as the Log4Shell Spring Boot example illustrates.
- Unsigned VEX from unknown sources. Third-party VEX for a product you did not get from that party is treated as authoritative.
Framework mapping
| Framework | Reference | Relevance |
|---|---|---|
| NIST SP 800-53 Rev. 5 | RA-5 Vulnerability Monitoring and Scanning; SI-2 Flaw Remediation; RA-3 Risk Assessment | Using VEX to validate and prioritise scanner findings |
| NIST SP 800-53 Rev. 5 | SR-3 Supply Chain Controls and Processes; SR-4 Provenance; SA-15 Development Process, Standards, and Tools | Requiring and producing supplier vulnerability status information |
| NIST CSF 2.0 | ID.RA-01, ID.RA-08, GV.SC-05, GV.SC-07, PR.PS-02 | Validating vulnerabilities, handling disclosures, supplier requirements and risk |
| ISO/IEC 27001:2022 Annex A | A.8.8 Management of technical vulnerabilities; A.5.21 Managing information security in the ICT supply chain | Vulnerability evaluation and supplier information flows |
| MITRE ATT&CK v20 | T1195.001, T1195.002 | Supply chain compromise, including of dependencies, that VEX producers and consumers must guard against |
Practitioner checklist
- Decide which VEX format you will produce or accept, based on your SBOM ecosystem and customers.
- Make sure VEX product identifiers match SBOM identifiers exactly, including versions.
- Require a justification or impact statement for every not affected statement, and keep the evidence behind it.
- Publish under investigation early for high-profile CVEs and update as analysis completes.
- Sign VEX documents and protect the signing identity like a release key.
- Verify the author and signature of every VEX document before applying it.
- Suppress findings with a recorded reference to the VEX statement, never delete them.
- Check conditional justifications against your actual configuration and environment.
- Process VEX updates continuously and re-open findings when status changes to affected.
Standards and references
- OASIS: Common Security Advisory Framework Version 2.0
- OpenVEX: OpenVEX Specification
- CycloneDX: CycloneDX Specification and Schema
- SPDX: SPDX 3.0.1 Security Profile
- CISA: When to Issue VEX Information
- OpenVEX: vexctl
- Aqua Security: Trivy VEX documentation
Related NHI Mgmt Group resources: SBOM Guide · Reachability Analysis Guide · PSIRT Guide