Design a CVE intelligence feed
Design the internal service that ingests the CVE List, NVD, CISA KEV, EPSS and package advisories, merges them into one record per vulnerability with provenance, keeps working when a source lags or breaks, and turns severity, exploitation and exposure into a priority that teams can act on.
01Understand the problem and size it
- Know what each source contributes: CVE List (identity), NVD (enrichment), KEV (known exploitation), EPSS (likelihood), OSV and vendor advisories (affected versions)
- Ingest incrementally within each source's rate limits and survive lag and backlog
- Merge into one record with per-field provenance and history of what you knew when
- Prioritise with exploitation and exposure first, CVSS second
What they probe: whether you know CVSS measures severity rather than likelihood, how you poll NVD without being throttled, what you do when NVD has not scored a CVE yet, how a bad upstream update is stopped from wiping your data, and how priority combines with asset exposure. Where candidates slip: sorting by CVSS alone, treating NVD as the only source, no plan for rejected CVEs or upstream outages.
Every vulnerability management programme needs the same reference data: which vulnerabilities exist, what they affect, how bad they are and whether anyone is exploiting them. That data comes from several public sources with different owners, formats, update rates and failure modes. The feed's job is to make them one dependable internal source that scanners, ticketing and dashboards can rely on.
| Source | Owner | What it adds | Update pattern |
|---|---|---|---|
| CVE List (CVE JSON 5 records) | CVE Program; records written by CNAs | The ID, description, references, CNA-supplied affected products and often CVSS | Continuous; bulk and delta files published on GitHub |
| NVD | NIST | CVSS, CWE and CPE applicability statements | API 2.0, incremental by last-modified date; has carried a large enrichment backlog since early 2024 |
| CISA KEV catalogue | CISA | CVEs with evidence of active exploitation, with dateAdded, dueDate, requiredAction, knownRansomwareCampaignUse | A JSON and CSV feed, changed a few times a week |
| CISA Vulnrichment (ADP) | CISA | SSVC decision points, CVSS and CWE added to CVE records | Inside CVE records |
| EPSS | FIRST | Probability of exploitation activity in the next 30 days, and its percentile | Daily file of scores for all published CVEs (version 4 since 2025) |
| OSV, GitHub Advisory Database | Open source communities | Exact affected package versions by ecosystem | Continuous |
Requirements
- Ingest all sources incrementally and keep a merged record per CVE with the origin of every field.
- Publish changes as events (
cve.updated) and expose a query API to internal consumers. - Compute a priority per CVE, and per CVE-asset pair when given exposure.
- Keep history: what the record looked like on any past date, for audits and SLA disputes.
CVE records total roughly 300,000, growing by over 40,000 a year (~110 / day, bursty)
NVD API page size up to 2,000 CVEs per request
NVD rate limit 5 requests / rolling 30 s without a key, 50 with a key
full backfill 300,000 / 2,000 = 150 requests
with a key 150 / 50 x 30 s = 90 s
without a key 150 / 5 x 30 s = 900 s = 15 min
daily EPSS file one row per published CVE, a few MB compressed
KEV well over a thousand entries; a small file
merged record ~10 kB with history -> 300,000 x 10 kB = 3 GB (one database)This is an advanced module
Design problems: vulnerability and exposure is part of the Pro pass. 4 more sections — the deep dives, the final design, follow-ups and the test — unlock with it.
Already a cyart.io or studio.cyart.io member? Your pass code is in your Studio account.
Related lessons
- CISA KEV and what "known exploited" meansAnalyst · Vulnerability triage & CVE prioritisationFree
- Building a patch-priority queueAnalyst · Vulnerability triage & CVE prioritisationFree
- EPSS and exploit likelihoodAnalyst · Vulnerability triage & CVE prioritisationFree
- Design a vulnerability management platformHLD · Design problems: Infrastructure & dataPro