Skip to content

Sources and evidence

Every fact BreachSpider shows links to the document that states it, with a date that says what it means: when we read it, the source's as of date, when it was retrieved, or when the source says it was published or modified. Naming a source here describes where data comes from; it is not an endorsement by that source.

Vulnerability sources

Source What it gives How often
CVE records from the National Vulnerability Database (NVD) Descriptions, severity scores, product and version configurations (CPE) Nightly (02:00 UTC), with a nightly completeness check against the published record list
CVE Program records (the CVE numbering authorities' own records, including their ADP enrichment) Affected and fixed versions in the vendor's own words Weekly
Vendor security advisories in the CSAF format (for example Siemens, Schneider Electric, ABB, SICK, CODESYS, Phoenix Contact, Rockwell Automation, Beckhoff) Precise affected ranges, fixed versions, "no fix" statements, mitigations Imported in batches, each through the integrity gate
Government ICS advisories (CSAF documents published by CISA) Restated vendor ranges and fixes, the advisory link Weekly (Saturdays), advisory map refreshed Mondays
Microsoft security update data The fixed build and update (KB) for every CVE on every Windows build line Monthly after Patch Tuesday, plus a daily change check
Vendor PSIRT pages Advisories from vendor security teams Daily
National CERT feeds (for example JPCERT, JVN, ANSSI, CCCS, CERT-EU) Advisories and references Nightly

As of 2026-11-08 BreachSpider holds about 401,000 CVE records (rejected IDs excluded), 3.4 million affected version rows, of which about 229,000 carry a vendor, CVE numbering authority or government advisory provenance, and 1.5 million links between CVEs and 169,000 catalog products from 36,000 vendors.

Exploitation signals

Source What it gives
The Known Exploited Vulnerabilities catalog (published by CISA) Whether a CVE is known to be exploited, the date it was listed, the catalog's due date and whether ransomware use is known. Read daily at 03:30 UTC. About 1,769 entries on 2026-11-08
Exploit-DB Public exploit code entries. Read daily
A public exploit repository index Repositories that claim exploit code, "as of" the index commit we hold. Repository contents are not reviewed. Last refreshed 2026-05-23
EPSS (Exploit Prediction Scoring System) The probability a CVE is exploited in the next 30 days, and its percentile

Catalog due dates

The known-exploited catalog's due date is the date set for US federal agencies. BreachSpider shows it because it is a useful urgency signal; it is not a deadline set for your organization.

Lifecycle and firmware sources

Support status comes only from vendor statements: lifecycle pages, end-of-sale and end-of-life notices, discontinued-product lists, and end-of-life statements inside the vendor's own security advisories. Every date links to the page it came from, with the day it was checked. A third-party listing (for example the community-maintained endoflife.date) is always labelled as such and always yields to a vendor fact of the same kind. See Lifecycle and firmware currency.

Source tiers

When sources disagree, BreachSpider prefers:

  1. Tier A, the vendor's own statement (the vendor's advisory, CSAF document, lifecycle page, or the vendor's own CVE record as numbering authority).
  2. Tier B, a government restatement of the vendor's statement.
  3. Tier C, a third-party listing, always labelled, never presented as the vendor's word.

A vendor's stated fixed version outranks a version derived from a range end. A vendor "no fix" statement is quoted in the vendor's words.

Collection rules

  • Each connector reads the site's robots.txt and stores it with the terms page and a checksum before collecting.
  • A refusal (HTTP 403 or 429) stops the run at once and is never worked around.
  • Nothing behind a login is read.
  • Only facts with a link are kept. Vendor material is not republished.
  • Where a vendor's website terms ban automated collection, statements are transcribed by hand with exact quotes and the page link, and that site is never fetched by a connector.

How data changes are checked

Every data change goes through an integrity gate: it is applied inside one database transaction, checked against critical invariants (golden device sets, Windows results reproduced exactly, provenance on every new range) and committed only on a pass. A catalog audit tests product identity and version boundaries against the vendors' own ranges and is re-run nightly.

In the product

  • The CVE record in the web app has an evidence table and a JSON evidence download.
  • The API returns references.vendor_advisories, references.cisa_ics_advisories, references.cve_org_url and, per CVE on a device, affected_range with its source. See Correlate CVEs.
  • Reports record the data date of every source they used.