CVE Census

Open data · as of 16 Nov 2026

Take the joined data with you

The tables behind every figure, rebuilt every day. Parquet for analysis (DuckDB, pandas, Polars, R, Spark), CSV for everything else.

TableRowsDownloadWhat it holds
cves407 052Parquet 43.5 MB
CSV 50.2 MB
One row per published CVE: dates, CNA, scores, weakness, NVD status, exploitation and exploit facts, EPSS, and counts from every source.
cve_events1 503 349Parquet 2.8 MB
CSV 5.4 MB
One row per event in a CVE’s life: (cve_id, event, date, source).
credits123 454Parquet 830 kB
CSV 1.0 MB
Who is credited for which CVE, as written in the record, and in which role.
cve_vendors470 002Parquet 2.9 MB
CSV 3.4 MB
Vendors and products named in each CVE record (vendor names merged on case and company suffix).
cnas615Parquet 59 kB
CSV 45 kB
Every CNA: the CVE Program’s list joined with how many records each has published.

Every CVE as JSON

Each CVE page reads a small JSON file that you can use too: /data/cve/<year>/<number ÷ 100>.json holds 100 IDs. For example, CVE-2021-44228 is in /data/cve/2021/442.json. Every CNA, vendor, researcher and weakness has its own file under /data/cna/, /data/vendor/, /data/researcher/ and /data/cwe/, named like its page.

Querying it

-- DuckDB, straight from the site
SELECT cna, count(*) AS exploited
FROM 'https://cve.webcracy.ch/data/open/cves.parquet'
WHERE exploited GROUP BY 1 ORDER BY 2 DESC LIMIT 10;

Licence and credit

The census's own work (the joins, the derived fields, the figures) is free to use with credit to “CVE Census, Webcracy.ch”. The underlying data stays under its sources' terms; see sources. VulnCheck data, when used, appears only in totals, as its terms require.