← CRA Glossary
Software Supply Chain

Open Source Vulnerability (OSV) Format

OSV (Open Source Vulnerability) is a standardised vulnerability data format and database maintained by Google, designed for precise matching of vulnerabilities against specific package versions in open source ecosystems. It provides a PURL-based alternative to NVD for open source vulnerability correlation in SBOMs.

OSV (Open Source Vulnerability) is a standardised vulnerability data format and database maintained by Google, designed for precise matching of vulnerabilities against specific package versions in open source ecosystems. It provides a PURL-based alternative to NVD for open source vulnerability correlation in SBOMs.

Software Supply Chain

Last updated 19 July 2026

What Is the OSV Format?

The Open Source Vulnerability (OSV) format is a standardised schema for representing vulnerability information, maintained by Google and hosted at osv.dev. OSV records describe which package versions are affected by a vulnerability with ecosystem-specific precision - specifying the exact version ranges in which the vulnerability is present and the first fixed version. This version-range model is more precise than the NVD CPE-based approach, which often has known inaccuracies for package-level matching. OSV data is provided for major open source ecosystems including npm, PyPI, Maven, Go, Rust (crates.io), Ruby Gems, NuGet, Debian, Alpine, and Android. The OSV schema is open and allows vulnerability databases from multiple sources to be published in a consistent, machine-readable format.

CRA reference:Annex VII

OSV for SBOM-Based CRA Compliance

For manufacturers building CRA-compliant vulnerability management workflows, OSV provides a powerful complement to NVD/CVE data. The key advantage is precision: an OSV query for pkg:pypi/requests@2.28.0 returns only the vulnerabilities that specifically affect that version, using the OSV version-range data. NVD CPE-based queries for the same package often return false positives (CVEs with imprecise CPE data) and false negatives (CVEs missing CPE data). OSV-based vulnerability correlation against SBOM PURLs provides:

  • Higher signal-to-noise ratio in vulnerability reports.
  • Version-specific 'affected' vs 'fixed' information enabling rapid determination of whether an available fix resolves the vulnerability.
  • Data aggregated from multiple sources (OSV, GitHub Advisory Database, GHSA) in a single query.
CRA reference:Annex VII

OSV-Scanner and SBOM Integration

Google provides OSV-Scanner, an open-source command-line tool that takes an SBOM (CycloneDX or SPDX) or a lockfile as input and queries the OSV database for known vulnerabilities affecting the listed components. OSV-Scanner is designed to integrate into CI/CD pipelines for continuous vulnerability monitoring. For CRA compliance workflows:

  • Run OSV-Scanner against the product SBOM on every build to detect new vulnerabilities disclosed since the previous scan.
  • Include OSV-Scanner results in the technical documentation file as evidence of vulnerability monitoring.
  • Use OSV-Scanner output to feed the vulnerability triage process, prioritising findings by CVSS score and EPSS data.
  • Generate VEX statements for OSV findings that are assessed as not-affected in the product context.

OSV Relationship to GitHub Advisory Database and CVE

OSV serves as an aggregator and re-publisher of vulnerability data from multiple sources. GitHub Advisory Database (GHSA) is the primary source for many npm, PyPI, and Go vulnerabilities. OSV imports GHSA data and republishes it in OSV format. CVE records from NVD are also mapped to OSV entries where precise version data is available. This means an OSV query may return both OSV-native records (e.g., GHSA-xxxx-xxxx-xxxx) and CVE-mapped records (e.g., CVE-2023-XXXXX), potentially providing more complete coverage than querying either source alone. Manufacturers should include OSV-based scanning alongside NVD-based scanning in their SBOM vulnerability management workflow to maximise coverage.

Work through the CRA for each product in one workspace.

Classify each product, build the Annex VII technical file and prepare Article 14 notifications. The disclosure portal that receives vulnerability reports stays free.

Frequently asked

Is OSV coverage as good as NVD for all vulnerability types?+

OSV is strongest for open source software packages distributed through public registries (npm, PyPI, Maven, Go, etc.). NVD has broader coverage including proprietary software, hardware, and OS-level vulnerabilities. For a typical CRA product with a significant open source component stack, OSV + NVD together provide better coverage than either alone. Manufacturers of products with significant open source content should prioritise OSV for package-level scanning. NVD remains essential for coverage of closed-source, OS, and hardware components.

How often is the OSV database updated?+

The OSV database is updated continuously - multiple times per day - as new vulnerability records are published or updated in source databases like GitHub Advisory Database and NVD. OSV-Scanner fetches current data at scan time, meaning each scan reflects the latest known vulnerabilities. For CRA continuous monitoring purposes, running OSV-Scanner scans daily (automated in CI/CD) provides near-real-time alerting for new vulnerabilities affecting the product's SBOM.

Can OSV be used for vulnerabilities in firmware or embedded OS components?+

OSV's coverage is strongest for high-level application packages. Coverage for firmware-specific components, embedded RTOS packages, and hardware-related vulnerabilities is more limited. For embedded products, OSV should be supplemented with NVD scanning for OS-level packages (using CPE or PURL for deb/rpm/apk packages) and ICS-CERT advisories for firmware-specific vulnerabilities. Tools like Grype (Anchore) support both OSV and NVD sources in a single scan and handle container/OS package types well.

Sector checklists covering Open Source Vulnerability (OSV) Format

Consumer Routers & ModemsConsumer routers and modems are high-value targets for attackers and face specific CRA requirements around default credentials, remote management security, and firmware update integrity. Routers, internet modems, switches and wireless access points are Important Class I (Annex III, point 12), for home, industrial and critical infrastructure markets alike. No CRA harmonised standard is cited yet, so today a router needs a notified body for conformity assessment (Article 32(2)).Enterprise Networking EquipmentEnterprise networking equipment - switches, firewalls, load balancers, and network management systems - spans multiple Annex III classifications. Firewalls in hardware or software form and intrusion detection and prevention systems are Important Class II. Switches, routers, wireless access points, network management systems and SIEM systems are Important Class I. A load balancer is default class unless its core functionality is a listed category, such as a web application firewall. Class II products always need third-party conformity assessment (Article 32(3)). Class I products need a notified body today, because no CRA harmonised standard is cited yet (Article 32(2)).Embedded Linux DevicesEmbedded Linux devices span a vast range - from network gateways and industrial HMIs to set-top boxes and smart displays. All are products with digital elements subject to the CRA. The open-source nature of Linux creates specific obligations around SBOM completeness and CVE monitoring. The class follows the core functionality of each device (Article 7(1)). An industrial HMI is default class. A network gateway whose core functionality is a router is Important Class I (Annex III, point 12). Industrial or infrastructure use does not change the class.

This definition is part of the EU Cyber Resilience Act guide, which explains Regulation (EU) 2024/2847 article by article.

Browse the full CRA Compliance Checklist

See how Open Source Vulnerability (OSV) Format fits into your complete CRA compliance programme.

View checklists →