Skip to content
EPSS vs CVSS

COMPARISON, read 10 Nov 2026

KEV confirms exploitation after it starts; CVSS exists at publication but cannot tell you who is already being attacked

All comparisonsThe ledger

Maintained by
CVSS: FIRST. KEV: CISA
Scale
CVSS: 0.0 to 10.0. KEV: listed or not listed
Current version
CVSS v3.1 and v4.0. KEV catalog 2026.11.10, 1,779 entries
Updated
CVSS: on scoring. KEV: on each addition
Answers
CVSS: the severity of the flaw. KEV: whether exploitation is confirmed

Different evidence, different timing

A CVSS score is available when a CVE is scored, usually near publication, and rests on the flaw's characteristics. A KEV entry exists only after CISA holds reliable evidence of exploitation. BOD 22-01, the original KEV directive and now superseded, stated the inclusion test as reliable evidence that the exploit is being actively used to exploit public or private organizations by a threat actor.

The result is a trade. CVSS is early and broad but silent on attackers. KEV is late and narrow but states a fact. The KEV Lag Ledger measures how late: the median gap from NVD publication to KEV listing was 10.5 days across 318 additions, with 144 (45.3 percent) listed within 7 days and 192 (60.4 percent) within 30.

PropertyCVSSKEV
Evidence typeFlaw characteristicsObserved exploitation
AvailableNear publicationAfter CISA confirms
CoverageNearly every CVE1,779 entries on 2026-11-10
ReproducibleYes, formula or lookupYes, public JSON feed with catalogVersion
FitsSeverity floor, SLA by tierOverride, deadline, escalation

Where CVSS misses KEV

Of the 318 additions, 33 (10.4 percent) scored below 7.0 and one scored below 4.0. A program that scheduled only High and Critical CVSS findings would not have scheduled those 33 until KEV listed them. For them, KEV supplied the missing information.

Timing matters as much as score. The 13 low-scoring entries listed within 7 days of publication left a window of days. The three low-scoring entries listed more than 365 days after publication show the other side: an old Medium finding in a backlog that changed meaning on one date.

Where KEV misses CVSS

KEV does not list a flaw that is dangerous but not yet attacked, or attacked but not yet seen. The median CVSS across the additions is 8.8, and 146 of 318 (45.9 percent) are 9.0 or above, so CVSS scored this set high more often than not. The ledger holds only listed CVEs, so it cannot show how many other CVEs share those scores.

The ledger shows the lead time. For additions scored 9.0 or above the median gap to listing was 8.0 days; for those below 7.0 it was 27 days. A CVSS Critical finding in the first week is therefore a candidate for early action even though it is not in KEV. KEV is silent about it until CISA acts.

CVSS also varies by assessor. In the ledger, 130 of the 318 scores (40.9 percent) come from nvd@nist.gov and the rest from CNAs such as secure@microsoft.com (28) and psirt@cisco.com (16). The same flaw can carry a different score from different assessors, and the NVD announced on 2026-04-15 that it will no longer routinely add its own score where a CNA has supplied one. KEV has one maintainer and one inclusion test.

What the law asks

BOD 26-04, issued 2026-06-10, supersedes BOD 22-01 and BOD 19-02. It sets remediation urgency from asset exposure, KEV status, exploit automation and technical impact. CISA states that technical impact is similar to the CVSS base score's concept of severity. The CERT/CC page for the directive's decision tree lists four inputs and none is a CVSS score; a binary Technical Impact plays that role.

The consequence for a federal reader is that KEV is a required input and CVSS is a supporting one. The KEV feed shows the timelines at work: 70 of the 92 entries added from 2026-09-01 carry a 3 day due date and 22 a 14 day one.

The feed also carries a due date. For the 92 entries added from 2026-09-01 through 2026-11-10, computed from the dueDate and dateAdded fields, 70 fall due in 3 days and 22 in 14 days. CVSS has no deadline field, so a severity-based program writes its own service levels, while a KEV-based program receives them.

A combined rule

Put KEV membership first, as an override on any asset where the CVE is present. Place CVSS Critical and High below it, with the version and the vector string recorded. Add EPSS for the rest. Every item then has a stated reason: confirmed exploited, severe, or likely.

Record the date of the KEV check. The feed is a snapshot with a catalogVersion, so a report that says the asset had no KEV findings should name the version.

The ransomware field adds a second KEV attribute with no CVSS counterpart. 29 of the 318 additions are marked Known. They are a small share, and the flag is the only public attribute in these three signals that records use by ransomware operators. A program may set a faster service level for flagged entries.

What to put in the policy

Name both signals and their roles in one sentence: KEV determines whether the vulnerability is already being exploited, CVSS determines how severe it is if it is. Name the CVSS version. Name the KEV source and its refresh interval. State who may override and how the override is logged.

The ledger is the evidence for including both. KEV alone would have queued nothing until a listing. CVSS alone would have missed the 33.

Capacity planning follows from the monthly counts. KEV additions ran from 17 in 2026-01 to 43 in 2026-09 among full months, which is the load a KEV-first policy places on remediation teams. A CVSS-first policy places a load that depends on how many CVEs in your inventory score 7.0 or above, a number only your own data can give.

Verdict by situation

  • Federal agencies and contractors: Rank by KEV status first, because BOD 26-04 makes KEV membership a required input to the remediation timeline and CVSS only a supporting one.
  • Large enterprises with ten or more scanners: Rank by KEV first as an override on every asset tier, then by CVSS within the remaining queue, so the confirmed-exploited set never waits behind a severity sort.
  • Managed security providers: Rank by KEV first because it is one public list that every client can verify, and use CVSS to order each client's backlog.
  • Mid-size teams replacing Kenna or a spreadsheet: Rank by KEV first because a small team cannot argue with a confirmed-exploitation list, and use CVSS Critical and High only to fill the remaining capacity.
  • Application security and infrastructure in one queue: Rank infrastructure CVEs by KEV first and application findings by CVSS first, because KEV lists only CVEs and many code-level findings carry none.

Other comparisons

Sources

  1. FIRST, CVSS v3.1 specification, read 10 Nov 2026
  2. CISA, KEV catalog JSON feed (catalogVersion 2026.11.10), read 10 Nov 2026
  3. CISA, BOD 26-04, archived copy, read 10 Nov 2026
  4. CISA, BOD 22-01 (superseded), archived copy, read 10 Nov 2026
  5. CERT/CC, CISA BOD 26-04 Response Model, read 10 Nov 2026

Last reviewed 10 Nov 2026.