Sorting CISA's KEV Catalog by CVE Year Before You Book a Pen Test
If you run IT for a small company and somebody asks you for a penetration test, I'd spend an afternoon with CISA's Known Exploited Vulnerabilities catalog before you call a single vendor. We did that for our Dallas clien
If you run IT for a small company and somebody asks you for a penetration test, I'd spend an afternoon with CISA's Known Exploited Vulnerabilities catalog before you call a single vendor. We did that for our Dallas clients in November 2026, sorting a year of entries by the year in each CVE number, and what we found is the reason we now tell almost everyone to scan before they test.
Below is the code we used, what came out of it, and the scoping problems we keep running into once a test actually gets booked.
Scanning and testing are separate jobs
NIST SP 800-115, the Technical Guide to Information Security Testing and Assessment, treats them as two different techniques. You'll find vulnerability scanning in the chapter about finding and identifying targets, while penetration testing has its own chapter about validating vulnerabilities, where NIST walks through 4 phases from planning and discovery to the attack and the final report.
When you run a scanner, it compares your devices, servers and websites with databases of published flaws, and you get back a ranked list of things to patch. When you hire a tester, they get a signed scope and try to use those flaws, plus any weak passwords and loose permissions they find, to show you how far an intruder could get. You can rerun the scanner every month for very little, but you'll pay the tester by the day.
Pulling the catalog
CISA publishes the KEV catalog as a JSON feed, and every entry in it carries a cveID, a vendorProject, a product and a dateAdded. On November 16, 2026 we downloaded version 2026.11.13, which had 1,781 entries in total. Then we kept the 12 months before the catalog date and counted entries by the year inside each CVE number
const k = JSON.parse(require('fs').readFileSync('kev.json', 'utf8'));const v = k.vulnerabilities.filter(x => x.dateAdded >= '2025-11-14' && x.dateAdded <= '2026-11-13');const yr = x => +x.cveID.split('-')[1];const byYear = {};
v.forEach(x => byYear[yr(x)] = (byYear[yr(x)] || 0) + 1);
console.log(k.catalogVersion, v.length, byYear);
console.log('2023 or earlier', v.filter(x => yr(x) <= 2023).length);
In that 12-month window CISA added 322 entries. You'd expect most of them to be new, and they were, with 209 carrying a 2026 CVE number and 58 a 2025 number. What surprised the people we showed it to was the rest of the list. According to the catalog, 45 of the 322, or about 14%, had CVE numbers from 2023 or earlier, 22 were from 2020 or earlier, and the oldest one dated to 2008.
Every one of those old entries already has a patch available, which means a basic scan would point them out the first time you ran it. If you book a tester before cleaning them up, you end up paying a daily rate for a report that lists missing patches.
Vendors and edge devices
When we grouped the same 322 entries by vendorProject, Microsoft came out with 42, Cisco with 22 and Fortinet with 12. We also counted 90 entries, about 28%, in firewalls, VPN gateways, remote access tools and file transfer servers from vendors such as Fortinet, Ivanti, SonicWall, Citrix and Palo Alto Networks. Because those devices sit on the internet, your external scan sees them exactly the way an attacker does, and we'd make sure that scan is right before worrying about anything else.
Verizon's 2026 Data Breach Investigations Report backs this up. It found that exploited vulnerabilities started 31% of breaches, more than any other entry point, while credential abuse started 13%. A scanner is good at the first group and can't do anything about the second.
How we set up the scans now
After going through the catalog, this is the plan we hand to clients.
Build the inventory first, and include the old machines in the back room and the laptops that only ever connect from home, since a scanner won't flag a device it doesn't know exists.
Scan from outside and inside at least once a quarter, or every month if somebody has time to work the fixes. PCI DSS already sets quarterly as the minimum for card data.
Patch in KEV order, so anything that shows up in the catalog goes to the top of the fix list.
Book the penetration test once the scan results are mostly clean, so the tester's hours go to things a scanner can't see.
Have the tester retest your fixes, and file the retest letter with the report.
Gotchas we hit when the test gets booked
Plant networks. Some of the manufacturers we work with run production machines on the same network as the office. We scan those networks with a list of controllers to leave alone, and we test the office side after the shift ends.
Stolen credentials. According to the City of Dallas after-action report, the city finished a 2-week penetration test with CISA in February 2023, and Royal ransomware still got in on April 7 through a service account. You won't see a stolen service account password in any scan report, so put service accounts in the test scope and treat sign-in monitoring as a separate control.
The authorization letter. Somebody who owns the systems has to sign it. If a cloud provider or your landlord owns part of the network, you may need their approval too, and it's better to ask early.
Independence. When the company that manages your network also offers to test it, ask how they keep the tester separate from the people who built it.
Network changes. If you move offices or roll out a new line-of-business system, the network has changed enough that we'd scan again before anybody tests.
Checking the price. At the $155.78 median federal hourly rate for penetration testers, a $1,000 quote covers about 6 hours. Ask for a sample report too. When it reads like a spreadsheet of CVE numbers with severity colors, you're looking at scanner output, whereas a real report reads like a sequence of steps from the first foothold to the furthest point the tester reached.
Takeaway
Run the KEV feed against your own inventory before you spend money on tester days. The old, patchable flaws it turns up are cheap to fix, and otherwise they'd take up part of the engagement you're paying for.
We put the full comparison for Dallas businesses, with the rules table and sources, in our guide to penetration test vs vulnerability scan decisions.
Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.