CPE Review¶
SentriKat matches your software against vulnerability data through CPE identifiers (the naming standard NVD uses for products). Getting this mapping right is the difference between a trustworthy dashboard and alert noise — so SentriKat treats it with a simple, strict philosophy:
A wrong CPE is worse than no CPE.
How assignment works¶
- Exact matches first. Software reported by agents or imports is matched against the local NVD CPE dictionary (tens of thousands of vendor:product pairs, synced automatically). Exact dictionary matches win over everything else.
- A strict plausibility gate for everything fuzzy. Near-matches must pass checks on BOTH the vendor and the product name; generic words (client, agent, runtime, setup, redistributable…) and substring tricks don't get through. Installers and sub-components never inherit the parent product's CPE.
- Below the bar → NO CPE, and a review entry. Instead of guessing, SentriKat queues the software on the CPE Review page for a human decision. Deterministic, reproducible — no AI in the matching engine.
The same rules re-validate existing assignments: a weekly self-audit flags entries that diverge from the NVD dictionary and can self-heal wrong ones.
The CPE Review page¶
Products → CPE Review lists software that didn't clear the bar, grouped by proposed identity:
- Assign — confirm a proposed CPE (it becomes a learned mapping; matching re-runs immediately for that product)
- Dismiss — mark the software as not-matchable (e.g. an internal tool); per-row or for a whole group at once
- Learned mappings are remembered and, if you opted in, shared with the SentriKat knowledge base so every install benefits
What this means for your dashboard¶
- Matches shown by default are version-verified where the data allows it; unverifiable legacy matches are hidden behind an explicit filter instead of padding your counts
- 0-day / actively-exploited entries are always surfaced, flagged distinctly on the dashboard
- The "Needs Verify" card counts matches waiting for verification — worth a look on day one, near-zero in steady state
FAQ¶
Why does some software show "no CPE"? Because nothing cleared the plausibility bar — by design. Assign it in CPE Review if you know the right identity, or dismiss it if it's not a publicly tracked product.
Do I have to maintain mappings by hand? No. Exact dictionary matching + the review queue for edge cases is the whole workflow; there are no mapping files to maintain.