Independent guide to Regulation (EU) 2024/2847 · Status: in force
Analysis · the CRA vulnerability-handling loop

Vulnerability analysis & SBOM

The CRA turns your software bill of materials into a living obligation: know what is in your product, check it for open vulnerabilities, and keep watching across the support period so you can fix and report within Article 14's deadlines. Three steps, on one page.

1→

Generate a compliant SBOM

Produce a machine-readable bill of materials covering at least your top-level dependencies, with the right tool for how your product is built.

How to generate it ↓
2→

Assess & fix today

Run that SBOM against live vulnerability data now, find anything already open, and fix it before placing the product on the market.

Assess your SBOM ↓
3

Automate monitoring

Set up continuous monitoring and reporting so new vulnerabilities are caught inside the 24-hour window, not in a weekly digest.

Build the pipeline ↓
A continuous loop across the product's support period
Step 1

Generate a compliant SBOM

A software bill of materials is a hard legal requirement under Annex I, Part II, Point (1). This covers the mandatory scope and format, the tool to use for how your product is built, and the full lifecycle guide.

1 · Legal basis

Annex I, Part II ("Vulnerability Handling Requirements"), Point (1) requires manufacturers to "identify and document vulnerabilities and components … including by drawing up a software bill of materials in a commonly used and machine-readable format covering at least the top-level dependencies of the product."

Unlike some CRA provisions that allow interpretive flexibility, the obligation to produce a machine-readable SBOM covering at least top-level dependencies does not permit alternative approaches.

2 · Mandatory scope and format

Component coverage. The SBOM must document all top-level dependencies, the legal floor rather than the recommended target. Map transitive (indirect) dependencies wherever feasible; a shallow SBOM satisfies the letter of the law but is often insufficient for effective vulnerability management.

Format. The CRA mandates a "commonly used, machine-readable format". Accepted formats include:

  • SPDX (ISO/IEC 5962:2021): widely adopted, licence-focused, suited to compliance documentation.
  • CycloneDX: security-centric, well-suited to vulnerability-management workflows.
  • Other structured formats (JSON, XML) from recognised tooling are generally acceptable under the baseline requirement.
Important

Do not store SBOMs as PDF. PDF may be challenged as non-machine-readable by market surveillance authorities. Store the native JSON, XML or tag-value output from your toolchain.

Required metadata fields

FieldDescription
Component nameUnique identifier for the library, package or module
VersionExact version string; version ranges are insufficient
Supplier / originPublisher, vendor or open-source project name
Dependency relationshipsDirect vs. transitive; dependency graph where feasible
Cryptographic hashSHA-256 or stronger integrity check per component
Licence identifierSPDX licence expression (e.g. Apache-2.0, MIT)

3 · Generate your SBOM: pick the right tool

Most platforms can produce an SBOM automatically at build time, at no extra cost. The right tool depends on how your product is built.

Application and package ecosystems

  • GitHub / Actions: Dependency graph → Export SBOM (Settings → Code security). Outputs SPDX JSON.
  • GitLab: CycloneDX reports are generated natively by the dependency-scanning CI/CD job.
  • Syft (Anchore): open-source CLI producing SPDX and CycloneDX for container images, filesystems and package manifests.
  • cdxgen: CycloneDX SBOMs across npm, Maven, pip, Go, Rust and more.

Embedded C/C++ firmware

Package-manifest scanners such as syft and cdxgen assume a dependency manifest, which firmware does not have: components are statically linked or renamed inside the chip vendor's SDK. For that case:

  • Gangmu (github.com/GANGMU-SBOM/gangmu ↗): an open-source (Apache-2.0) build-time generator built for embedded C/C++. It reads your source tree, compile_commands.json, linker maps and firmware binaries, identifies open-source components a vendor SDK has renamed or statically linked, and emits CycloneDX 1.6 and SPDX 3.0.1, which feed straight into the Analyzer in step 2. It runs entirely on your machine, with no telemetry. Maintained by Suzhou Qilijie Technology.
Using Gangmu well

Gangmu is strongest when you have the source tree; binary-only analysis is partial, especially across CPU architectures. Prebuilt vendor libraries it cannot identify are still listed, with their path and SHA-256 hash and flagged as unidentified, so you can resolve them yourself.

Its SBOM embeds relative directory paths and file hashes from your tree, so review it before uploading it here or sharing it. Gangmu is one option among several SBOM generators listed here and is not affiliated with this website; we reviewed the tool and its rule base and credit it on its merits, it is not a sponsor.

Vulnerability screening. Before finalising any SBOM, screen every component against known-vulnerability databases. Both GitHub's scanner and Syft integrate with Grype for this. Including a component with a known, already-patched vulnerability is a direct violation of Annex I and carries no interpretive flexibility.

Warning · known vulnerability = violation

If a patched version of a component exists, you must use it. There is no "risk-accepted" category under the CRA for resolved vulnerabilities. Act on every finding before placing the product on the market.

4 · Lifecycle maintenance and retention

An SBOM is a living artefact, updated continuously across the product's supported lifecycle to reflect patched components, new dependencies, removed end-of-life components and supply-chain changes. All versions of the technical documentation, including SBOMs, must be retained for at least 10 years from first market placement…

Read the full guide

Sections 4–9: lifecycle, penalties, BSI TR-03183-2 & the readiness checklist

The rest of the guide covers the 10-year retention duty, confidentiality and supply-chain disclosure, integration with Article 14 vulnerability reporting, monitoring frequency, Germany's stricter BSI rules, the fines (up to €15M or 2.5% of turnover), and a ready-to-use readiness checklist.

We'll add you to the CRA newsletter. Unsubscribe anytime. No spam. See our privacy policy.
Step 2

Assess your SBOM today & fix open vulnerabilities

Before you place the product on the market, run the SBOM you just generated against live vulnerability data to find anything already open, and fix it. The free Vulnerability Analyzer does exactly that.

CRA Vulnerability Analyzer

The Analyzer cross-references every component against the National Vulnerability Database (NVD) and the EU Vulnerability Database (EUVD), flags End-of-Life components, and generates a compliance report you can attach to your technical file.

Open the Analyzer ↗sbom.i46.cz · opens in a new tab
The engine · free, hosted on i46
1Bring the SBOM from step 1 (CycloneDX or SPDX) and an active-vulnerability list (e.g. from debsecan).
2Upload both files; drag and drop or browse.
3Components are cross-referenced against NVD and EUVD; EOL status is tracked.
4Download a Word report with risk assessments and EOL documentation.
Fix before you ship

Every open finding must be resolved before the product is placed on the market: there is no "risk-accepted" category for vulnerabilities that already have a patch. Use the report to document what you found and what you did. Then move to step 3 to keep watching, because the SBOM is a living artefact, not a one-off check.

Step 3

Set up the automated monitoring pipeline

A clean SBOM today says nothing about tomorrow. You have to watch its components continuously, so a newly disclosed vulnerability is caught inside Article 14's 24-hour window. Here is how GitHub's native tracking is set up, and why notification digests on their own do not satisfy the law.

Critical for CRA compliance

Activating standard GitHub notification digests does not satisfy your legal obligations under EU market law. Article 14 binds manufacturers to a strict 24-hour early-notification window for actively exploited vulnerabilities. Summary digests are a developer-awareness feature and cannot replace real-time regulatory incident routing.

Part 1 · Repository setup (maintainers)

So GitHub builds a dependency graph and flags known vulnerabilities against the GitHub Advisory Database:

  • Open the repository → Settings.
  • In the sidebar, under Security, select Advanced Security.
  • Find Dependabot alerts and click Enable.

Part 2 · Personal security alert digests

Watch the repository. On the project's main page, open the Watch menu → All Activity, or Custom with Security alerts checked.

Set your digest frequency. Profile → Settings → Notifications → Dependabot alerts, then adjust the email digest:

  • Daily digest: open alerts combined into one packet every 24 hours.
  • Weekly digest: open alerts combined into one packet once a week.

Part 3 · Operational reality & CRA alignment

Relying on batched digests exposes you to compliance gaps when deploying to the EU market. How GitHub configurations map to CRA expectations:

ConfigurationBehaviourCRA status
Weekly digestAggregated summary once a week✕ Non-compliant; misses the 24-hour window completely.
Daily digestAggregated summary every 24 hours⚠ High risk; a zero-day dropped just after a digest breaches the deadline before the next one arrives.
Real-time webhooksJSON payload immediately on disclosure● Compliance enabler; immediate machine-readable details let triage teams document exploitability and contact ENISA in time.
Back to the engine

Real-time signal is necessary but not sufficient; you still have to assess each component and produce evidence. Feed every change back into the Vulnerability Analyzer ↗ to turn your SBOM and vulnerability data into a CRA-ready report, then report actively exploited vulnerabilities to ENISA and the CSIRT on the 24h / 72h / 14-day timeline.