CycloneDX VEX for Embedded Firmware: a Worked Synthetic Example
Write a scoped CycloneDX VEX decision for embedded firmware: an unresolved in_triage record, a synthetic not_affected record with justification, and reassessment triggers.
By CenitiaUpdated
Fact-checked against the law as of Reviewed by Vladimír Vician on How we check
A VEX statement should explain whether a specific vulnerability affects a specific product configuration and what evidence supports that conclusion. It should not become a convenient way to hide scanner findings that the team has not investigated.
This tutorial uses CycloneDX VEX and two deliberately fictional issues, DEMO-VULN-001 and DEMO-VULN-002. Neither is a CVE, a real advisory or a finding about any published library. Download the synthetic VEX document, which contains the first issue, and use the triage record.
Start with identity and a question
Identify the firmware version and hash, the component identity and the vulnerability source. Confirm that the scanner's package match is meaningful. A shared library name is not enough to establish that the vulnerable implementation is present.
Then phrase the question you need to answer: “Does the affected parser exist and remain reachable in this released image under the supported and reasonably foreseeable configurations?” This is more useful than asking whether the severity score looks alarming.
Record the evidence you do not have. If the supplier cannot provide the relevant binary or build configuration, the uncertainty remains part of the decision.
Use the CycloneDX vocabulary, not another format's
In CycloneDX 1.7 the analysis of a vulnerability sits in vulnerabilities[].analysis. The 1.7 schema allows these values:
| Field | Allowed values |
|---|---|
state | resolved, resolved_with_pedigree, exploitable, in_triage, false_positive, not_affected |
justification | code_not_present, code_not_reachable, requires_configuration, requires_dependency, requires_environment, protected_by_compiler, protected_at_runtime, protected_at_perimeter, protected_by_mitigating_control |
response | can_not_fix, will_not_fix, update, rollback, workaround_available |
detail | Free text |
firstIssued, lastUpdated | Timestamps |
Other VEX formats use different words. CSAF and OpenVEX have statuses such as fixed and under_investigation; CycloneDX has neither. A record with "state": "fixed" fails validation against the CycloneDX schema. The nearest CycloneDX states are resolved and in_triage.
Choose a state that matches the investigation
Use in_triage while evidence is incomplete.
A not_affected conclusion needs a justification from the list above and evidence for it. “We have never seen an exploit” and “the device is usually behind a firewall” do not by themselves establish that vulnerable code cannot be exploited. protected_at_perimeter exists as a justification, but it describes a control you can show, not an assumption about where customers install the device.
A resolved conclusion needs an identified corrected release and validation of it. Do not mark the installed population resolved because a patch has been merged into a development branch.
Follow the first example through its evidence gap
The demo firmware contains a fictional demo-parser component. The synthetic issue DEMO-VULN-001 concerns an optional decoding path. The example has no compiler output and no firmware binary that would show the path is absent.
The downloadable VEX therefore records in_triage, with a detail explaining that a build configuration and binary review are still required. This example is intentionally unresolved. It shows a machine-readable decision that does not invent a reassuring “not affected” result.
| Investigation input | Status in the demo | What a real engineer would supply |
|---|---|---|
| Firmware identity | Fictional release 1.0.0 | Actual released image hash |
| Component match | Synthetic identity | Proven component provenance and version |
| Feature configuration | Assumed optional feature | Recorded effective build flags |
| Code presence | Unverified | Binary or build evidence reviewed by an engineer |
| Reachability | Unverified | Interface and configuration analysis, tests where appropriate |
The demo record has no source object, because the issue is fictional. A real record should name where the vulnerability was published.
See what a supported not_affected record looks like
Suppose the right-hand column of that table had been filled in for a second fictional issue, DEMO-VULN-002: the release build record shows that the optional decoder was excluded, and a reviewer has confirmed that its code is absent from the released image. The record would then look like this:
{
"id": "DEMO-VULN-002",
"description": "Fictional issue used only for an editorial example; not a real vulnerability or CVE.",
"analysis": {
"state": "not_affected",
"justification": "code_not_present",
"detail": "Synthetic example. Assumes the release build record shows the optional decoder was excluded and a reviewer confirmed that its code is absent from the released image. No such evidence exists for this demo."
},
"affects": [
{
"ref": "demo-parser"
}
]
}
This record is not in the downloadable file, and the evidence it refers to does not exist; it only shows the shape of a supported decision. Added to the vulnerabilities array of the demo document, it is valid against the CycloneDX 1.7 schema. In a real record, put the references to the build record and the review in detail or in your triage record, so that a later reader can find them.
When the state of a record changes, update state, justification and detail, increase the document version, and keep the earlier document so that a reviewer can see how and why the conclusion changed.
Validate references as well as JSON
The affects reference must identify the intended component. Check that every reference resolves and that the product identity matches the triage record. A schema validator checks allowed fields and values; it does not establish that your analysis is true.
We validated the downloadable document against the official 1.7 schema on 1 October 2026, with JSON Schema format checks disabled, and checked that its affects reference resolves to a component in the document. That supports the structure of the example only. It says nothing about exploitability or about the completeness of a real vulnerability review.
The demo is a VEX document, not an SBOM: it lists only the component the statement refers to and has no dependency graph. If your recipients follow Germany's BSI TR-03183-2, keep it that way. That guideline requires an SBOM to contain no vulnerability information and recommends CSAF, including its VEX profile, for distributing it.
Give decisions an expiry condition
List the events that reopen the assessment: a firmware rebuild, changed build flags, a new hardware variant, a changed interface, new upstream evidence or a broader vulnerability description. A not_affected statement without such conditions can outlive the assumptions that justified it.
Keep the reviewer, decision date, evidence references and distribution scope with the record. Decide what can be disclosed publicly without exposing credentials, customer configurations or unnecessary exploit detail.
Keep regulatory reporting separate
A VEX state is not, by itself, a CRA reporting decision. Assess reporting triggers through the incident response process, and do not defer a required assessment because a JSON field still reads in_triage.
The two are connected, though. The Commission guidance on the CRA (C/2026/6013, point 218) states that a vulnerability in a third-party component that cannot be exploited in your product, for example because the vulnerable code is not reachable, does not qualify as an actively exploited vulnerability contained in your product. A not_affected analysis backed by evidence is the record of that conclusion; an unsupported one is not.
CRA applicability and timing
Most Cyber Resilience Act obligations apply to products placed on the market from 11 December 2027; a product placed on the market earlier becomes subject to them only if it is substantially modified from that date (Article 69(2)). Article 14 reporting is already a current duty: it has applied since 11 September 2026 to every in-scope product, including products placed on the market before 11 December 2027 (Article 69(3)), and it continues after the support period ends. Check the product’s scope, your role and the transitional provisions. See the Commission page for manufacturers, the Commission page on reporting and the Commission guidance on the application of the Cyber Resilience Act (C/2026/6013).
Continue reading
Related guides
tutorial
Secure OTA Firmware Updates and the Evidence to Keep
Plan secure OTA test evidence for signatures, interrupted updates, recovery and downgrade controls against the CRA update requirements, with a small host demo.
6 min read
guide
CRA Scope for Connected Devices and Their Cloud Backends
Apply the CRA remote data processing tests to connected device architectures and record external cloud dependencies separately.
5 min read
guide
When a Firmware Update Needs New CE Evidence or Retesting
Assess firmware changes against existing CE evidence, notified body notification duties and the CRA substantial modification rules with a change impact record.
6 min read
tutorial
CRA Cybersecurity Risk Assessment for an IoT Gateway
Create an IoT gateway risk register that connects product boundaries, abuse cases, security controls and verifiable release evidence.
4 min read
Put this into practice
Free tools & references
- CRA Readiness CheckerScore your product against the Cyber Resilience Act essential requirements.Open tool →
- EU Directive SelectorDescribe your product and find which EU directives and regulations apply.Open tool →
New to the terminology? Browse the compliance glossary — plain-English, citation-backed definitions of every term above.