CVSS, EPSS, KEV & SSVC, Compared

A scanner reports thousands of vulnerabilities; a team can patch a fraction of them this week. Four public systems help decide which. They answer four different questions, and most prioritisation mistakes come from using one to answer another’s.

4systems, four different questions about a vulnerability
30days — the window an EPSS probability covers
3criteria for a CVE to enter CISA’s KEV catalog
36paths through CISA’s SSVC tree to 4 decisions

None of these four is a risk assessment, and each says so. The CVSS v4.0 specification describes its output as scores “indicating the severity of a vulnerability relative to other vulnerabilities”, and FIRST’s own FAQ is blunter: “CVSS-B Base scores are not risk.” The EPSS FAQ says “EPSS is not a complete risk score.” They are inputs. Risk appears only when you add your asset, its exposure and what it supports — which is what SSVC, and CISA’s newest directive, are built to do.

SystemQuestion it answersOutputPublished by
CVSS v4.0How severe is this flaw, intrinsically?0.0–10.0 plus a vector stringFIRST (CVSS SIG)
EPSSHow likely is it to be exploited in the next 30 days?Probability 0–1 plus percentile, dailyFIRST (EPSS SIG)
KEVIs it already being exploited?In the catalog or not, with a due dateCISA
SSVCWhat should we do about it?Track, Track*, Attend or ActCERT/CC at Carnegie Mellon SEI, with CISA

1 · CVSS v4.0: Severity

The Common Vulnerability Scoring System is owned by FIRST.Org; version 4.0 was first published on 1 November 2023. Its metrics fall into four groups:

GroupMetricsSet by
BaseExploitability: Attack Vector (AV), Attack Complexity (AC), Attack Requirements (AT), Privileges Required (PR), User Interaction (UI). Impact on the vulnerable system (VC, VI, VA) and on subsequent systems (SC, SI, SA)Vendor or analyst
ThreatExploit Maturity (E): Attacked, Proof-of-Concept or UnreportedConsumer
EnvironmentalConfidentiality, Integrity and Availability Requirements of your asset (CR, IR, AR), plus modified Base metrics for your deploymentConsumer
SupplementalSafety, Automatable, Recovery, Value Density, Vulnerability Response Effort, Provider Urgency — extra context that does not change the scoreEither

v4.0 makes the source of a number explicit. A score must be labelled CVSS-B (Base only), CVSS-BT (Base + Threat), CVSS-BE (Base + Environmental) or CVSS-BTE (all three). The specification notes that providers such as NVD “typically provide only the Base Scores” — so the number in your scanner is almost always CVSS-B: severity in general, not in your environment.

CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N  →  network, no privileges, no user interaction, full impact: Critical

Optional severity ratings: None 0.0, Low 0.1–3.9, Medium 4.0–6.9, High 7.0–8.9, Critical 9.0–10.0. v4.0 also retired two v3 Temporal metrics (Remediation Level, Report Confidence) and renamed Exploit Code Maturity to Exploit Maturity.

2 · EPSS: Probability of Exploitation

The Exploit Prediction Scoring System is “a data-driven machine-learning model that estimates the probability that a published CVE will be exploited in the wild in the next 30 days.” It is maintained by the EPSS Special Interest Group at FIRST, with scores generated by Empirical Security and published free, for every CVE, every day (shortly after 13:30 UTC).

  • Probability vs percentile — the probability is absolute; the percentile is relative. A CVE at the 90th percentile scores higher than 90% of all scored vulnerabilities, which can still be a low probability.
  • What “exploited” means — that EPSS’s data partners’ sensors detect and share exploitation activity, not that you will be breached.
  • What it does not measure — damage. The FAQ: EPSS “does not measure how much damage successful exploitation would cause”.
  • It moves — scores change daily and the model is retrained periodically (the daily CSV’s header names the model version), so record the date with any EPSS value you quote.

3 · KEV and BOD 26-04: Exploited Now

CISA’s Known Exploited Vulnerabilities (KEV) catalog lists vulnerabilities with evidence of real-world exploitation — about 1,790 entries in November 2026. A vulnerability is added only when it meets three criteria:

  1. It has an assigned CVE ID.
  2. There is reliable evidence that it has been actively exploited in the wild. Scanning, security research or a published proof of concept do not count on their own.
  3. There is a clear remediation action, such as a vendor-provided update.

▸ The rules changed in June 2026. BOD 22-01 (November 2021) gave federal agencies two weeks to fix a KEV entry (six months for CVEs from before 2021). On 10 June 2026 CISA revoked it and BOD 19-02 and replaced both with BOD 26-04, “Prioritizing Security Updates Based on Risk”. Deadlines now come from a table “informed by” SSVC, using four inputs: whether the asset is publicly exposed, KEV status, whether exploitation is automatable, and technical impact. CISA’s example: a publicly exposed KEV with total impact that is automatable gets a 3-day deadline. And because BOD 19-02 is gone, federal agencies are no longer required to prioritise by CVSS.

BOD 26-04 binds US federal civilian executive branch agencies only; CISA encourages everyone else to follow it. KEV inclusion criteria are unchanged, and the catalog’s due dates are now set from the new table.

4 · SSVC: From Facts to a Decision

Stakeholder-Specific Vulnerability Categorization was created in 2019 by Carnegie Mellon University’s Software Engineering Institute (CERT/CC) with CISA; CISA built its own decision tree in 2020 and published the guide in November 2022. Instead of a score it asks four questions, each with a small set of values:

Decision pointValues
ExploitationNone (no evidence, no public PoC) · Public PoC (e.g. in Metasploit or ExploitDB, or a well-known method) · Active (reliable, shared evidence of use in the wild)
AutomatableNo · Yes — can reconnaissance, weaponisation, delivery and exploitation all be reliably automated? Unauthenticated remote code execution is likely Yes
Technical impactPartial (limited control or disclosure; denial of service) · Total (full control of the component, or disclosure of all its information)
Mission & well-beingLow / Medium / High, combined from mission prevalence (Minimal, Support, Essential) and public well-being impact (Minimal, Material, Irreversible)
DecisionCISA’s meaning
TrackNo action needed now; keep tracking and reassess on new information. Remediate within standard update timelines.
Track*Has characteristics that may need closer monitoring for changes. Remediate within standard update timelines.
AttendNeeds attention from supervisory-level staff; may involve asking for help or publishing a notification. Remediate sooner than standard timelines.
ActNeeds supervisory and leadership attention; internal groups meet to decide and execute a response. Remediate as soon as possible.

Try it: CISA SSVC decision

Mission & well-being
SSVC decision

Implements CISA SSVC Guide (November 2022) Table 8 and all 36 rows of Table 9 exactly. Mitigation status is tracked by CISA but does not change the decision.

5 · Using All Four Together

Each answers one question; a triage pipeline asks them in order. Two hypothetical findings from the same scan show why CVSS alone ranks them backwards:

Finding 1: VPN gateway flawFinding 2: library flaw
AssetInternet-facing remote-access gateway — the series’ ransomware entry pointInternal batch server, no inbound access
CVSS-B8.7 HighCritical (9.0+)
KEVYes — ransomware crews using itNo
EPSS0.9+ (top percentile)Under 1%
SSVCActive · Yes · Total · High → ActNone · No · Total · Low → Track
BOD 26-04 styleExposed, KEV, automatable, total → daysNormal patch cycle
  1. KEV first — anything exploited in the wild and present in your estate, starting with internet-facing assets.
  2. Decide with SSVC (or a BOD 26-04-style table) using your exposure and mission — the parts no public score can know.
  3. Sort the non-KEV remainder by EPSS and by asset criticality.
  4. Use CVSS for what it is — a consistent description of technical severity, and an input to SSVC’s technical impact. Record which CVSS (B, BT, BE or BTE) you used.
  5. Roll the result up into the enterprise risk register, where “unpatched exploited vulnerabilities on exposed assets” is tracked as a key risk indicator.

Bottom line

  • Four questions — CVSS: how severe; EPSS: how likely in 30 days; KEV: exploited now; SSVC: what to do.
  • None is risk on its own — their own publishers say so; risk needs your exposure and mission.
  • Label CVSS — a scanner’s number is almost always CVSS-B.
  • BOD 22-01 is gone — since June 2026, BOD 26-04 sets KEV deadlines from exposure, automatability and impact, as fast as 3 days.
  • KEV, then SSVC, then EPSS, then CVSS — and report the residue as a KRI.

Sources & method. FIRST, CVSS v4.0 Specification Document (v1.0 published 1 November 2023; current document version 1.2) and CVSS v4.0 FAQ; FIRST, EPSS home page and FAQ; CISA, Known Exploited Vulnerabilities catalog and inclusion criteria; CISA, BOD 22-01 (3 November 2021, revoked) and BOD 26-04 with its implementation guidance (10 June 2026); CISA, Stakeholder-Specific Vulnerability Categorization Guide (November 2022), Tables 1–9; CMU SEI / CERT/CC SSVC. All read in November 2026; short quotations as marked. The two findings are hypothetical, not real CVEs. Original analysis; no third-party graphics reproduced.

← Back to Cyber security