Why CVSS Alone Doesn't Tell You What to Patch First (And How KEV + EPSS Changes Everything)

PatchVex — Wed, 01 Jul 2026 19:51:17 +0000

Your Nessus scan just finished. You have 4,847 findings. CVSS says 200 of them are Critical.

Where do you actually start?

If your answer is "sort by CVSS score and work from the top" — you are making the same mistake most security teams make. And it is costing you hours every week while leaving your most dangerous vulnerabilities unpatched.

This article isn't arguing that CVSS is bad. CVSS is an excellent severity metric. The problem is using it as the only prioritization metric.

Try VulnPilot — free and open source

pip install vulnpilot
vulnpilot update-feeds
vulnpilot analyze scan.csv

Runs locally. MIT licensed. Your scan data never leaves your machine.
github.com/PatchVex/vulnpilot

The problem with CVSS

CVSS (Common Vulnerability Scoring System) measures the theoretical severity of a vulnerability. It answers the question: how bad could this be in the worst case?

It does not answer: is anyone actually exploiting this right now?

A CVSS 9.8 vulnerability on an isolated internal test server with no internet access is less urgent than a CVSS 7.0 vulnerability on your public-facing web server that attackers are actively exploiting today.

CVSS cannot tell you which one to fix first. It was never designed to.

CVSS only	KEV + EPSS + CVSS
Severity score only	Real-world exploit intelligence
Manual cross-referencing	Automatic prioritization
Hours of triage per scan	Results in seconds
No exploit context	Confirmed exploitation data
Misses active threats	Flags what attackers use today

What CISA KEV actually means

The Cybersecurity and Infrastructure Security Agency maintains a catalog called the Known Exploited Vulnerabilities (KEV) list.

This is not a theoretical list. Every CVE on this list has been confirmed as actively exploited in the wild. Real attackers. Real infrastructure. Right now.

At the time of writing, the KEV catalog contains only a tiny percentage of all published CVEs — those confirmed to be actively exploited in the wild.

If your Nessus scan returns 4,847 findings and 19 of them match the KEV catalog — those 19 are your actual priority. Not the 200 that CVSS calls Critical.

What EPSS predicts

FIRST (Forum of Incident Response and Security Teams) publishes the Exploit Prediction Scoring System (EPSS).

EPSS uses machine learning trained on real-world exploitation data to estimate the probability that a CVE will be exploited in the next 30 days.

A CVE with a high EPSS score has a high probability of being exploited in the near term. A CVE with a very low EPSS score is unlikely to be exploited soon — regardless of its CVSS score.

This is far more actionable than severity alone.

A real example — Log4Shell

CVE-2021-44228 (Log4Shell) has:

CVSS: 10.0 (maximum)
KEV: Yes — confirmed exploited
EPSS: Very high exploitation probability

Now consider a CVSS 9.8 vulnerability in an obscure library with no public exploit code, no KEV entry, and a very low EPSS score.

On a CVSS-sorted list it sits above Log4Shell. In reality it should be deprioritized until the confirmed exploited vulnerabilities are patched.

CVSS cannot make this distinction. KEV and EPSS can.

The composite scoring approach

VulnPilot combines KEV, EPSS, CVSS, and scanner severity into a transparent composite score designed to prioritize real-world risk.

The current scoring model uses:

Signal	Weight	What it tells you
CISA KEV	40%	Actively exploited right now
FIRST EPSS	35%	Exploitation probability
CVSS	15%	Severity context
Scanner rating	10%	Additional context

Any finding confirmed in the KEV catalog scores a minimum of 75/100 regardless of other factors. KEV means patch immediately — no further analysis needed.

The weighting is intentionally transparent and may evolve based on community feedback.

How VulnPilot works

Nessus CSV (local only)
        │
        ▼
   VulnPilot CLI
        │
   ┌────┴────────────┐
   │  CISA KEV       │  ← downloaded from cisa.gov
   │  FIRST EPSS     │  ← downloaded from first.org
   │  CVSS           │  ← from your scan data
   └────┬────────────┘
        │
        ▼
  Composite Score
        │
        ▼
  Prioritized Report

Only public threat intelligence feeds are downloaded. Your vulnerability scan data never leaves your machine.

What this looks like in practice

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  VulnPilot by PatchVex — Vulnerability Prioritization
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Total findings   : 4,847
  KEV matches      : 19    ← fix these first
  EPSS >= 90%      : 31
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  #  Score   Priority      CVE              Finding
  1  100.0   CRITICAL NOW  CVE-2021-44228   Log4Shell ★KEV
  2  100.0   CRITICAL NOW  CVE-2023-34362   MOVEit SQLi ★KEV
  3  99.8    CRITICAL NOW  CVE-2020-1472    Zerologon ★KEV
  4  11.5    LOW           N/A              SSH Weak Ciphers

The SSH weak ciphers finding sits at the bottom where it belongs — no KEV match, low EPSS. On a CVSS sort it would consume time that should go toward patching actively exploited vulnerabilities.

The time cost of doing this manually

Most security teams spend hours per scan cycle doing this manually:

Export Nessus CSV
Open in Excel
Sort by CVSS
Manually check KEV catalog
Manually check EPSS scores
Build remediation priority list
Repeat next week

The data already exists. The challenge is bringing it together automatically.

The missing piece for many teams is automating this workflow — locally, privately, without uploading scan data to a third-party service.

Why local-first matters

Vulnerability scan results often contain internal hostnames, IP addresses, software inventories, and infrastructure details. Many organizations prohibit uploading this information to third-party services. A Nessus CSV is essentially a map of every weakness in your infrastructure.

VulnPilot performs all analysis locally on your machine. Only public threat feeds are downloaded. Nothing about your infrastructure is ever transmitted.

What VulnPilot doesn't do

Being honest about limitations builds trust:

It does not replace your vulnerability scanner
It does not eliminate false positives
It does not make business risk decisions for you
It does not tell you asset criticality — you still need to consider what each host does

It helps prioritize remediation using publicly available exploit intelligence. The rest of the decision still belongs to your team.

VulnPilot currently supports Nessus CSV exports, with additional scanner support planned in future releases.

Install and try it

pip install vulnpilot
vulnpilot update-feeds
vulnpilot analyze scan.csv

# Generate a shareable HTML report
vulnpilot analyze scan.csv --html report.html

Free and open source. MIT licensed. Python 3.10+.

CVSS was designed to measure severity — not remediation priority.

If you're still sorting thousands of findings by CVSS alone, you're doing far more work than necessary.

VulnPilot combines publicly available exploit intelligence to help you answer one simple question:

What should I patch first?

Links:

Feedback welcome — drop a comment below or open a GitHub issue.

References

Built by PatchVex — privacy-first security tools for DevSecOps and security teams.

DEV Community: PatchVex