# Public exploit code shows a flaw can be attacked, but only observed exploitation shows that it is

SCORING SYSTEM, read 10 Nov 2026

**The short answer.** Exploit-DB, Metasploit modules and CISA's Vulnrichment record different strengths of exploit evidence: public proof-of-concept code, a ready-to-run module, and an assessed exploitation status. CISA Vulnrichment adds SSVC Exploitation values of none, proof of concept or active. KEV lists confirmed exploitation, and 29 of 318 recent KEV additions carry a ransomware flag.

- **Maintained by:** OffSec (Exploit-DB); the Metasploit project (Metasploit Framework); CISA (Vulnrichment)
- **Scale:** Presence of public code; SSVC Exploitation: none, proof of concept, active
- **Current version:** Continuously updated archives
- **Updated:** Per submission or per CVE
- **Answers:** Whether public attack code exists and whether anyone is using it

## Three kinds of evidence

Exploit evidence comes in strengths. Exploit-DB states that it is a CVE compliant archive of public exploits and corresponding vulnerable software, maintained by OffSec as a non-profit public service, and it is a repository of exploits and proofs of concept rather than advisories. The Metasploit documentation describes a free, open-source Metasploit Framework. A module in it means the exploit has been written into a tool that can be run by a person with moderate skill.

CISA's Vulnrichment repository adds an assessed status. CISA takes each CVE through SSVC scoring and records Exploitation as none, proof of concept or active. These are three different claims about the same CVE, and a program should not merge them into one flag.

CISA states how Vulnrichment fits the CVE record: the results are pushed back into the CVE corpus through the Authorized Data Publisher program launched in 2024, and consumers who already read live CVE data need not track the repository. For a program, the SSVC Exploitation field is thus available through the same CVE feed as the CVSS and CWE data.

| Source | What a hit means | What a miss means |
| --- | --- | --- |
| Exploit-DB | A public exploit or proof of concept exists | Nothing: private and unpublished exploits do not appear |
| Metasploit module | A framework module exists for the CVE | Nothing about other tools or private code |
| Vulnrichment Exploitation | CISA assessed none, proof of concept or active | None found in the record for CVEs not yet assessed |
| CISA KEV | Reliable evidence of active exploitation, by CISA inclusion criteria | The CVE is not confirmed as exploited by CISA |

## How the scoring systems consume it

CVSS v4.0 has one threat input, Exploit Maturity, based on exploit techniques, exploit code availability or active in-the-wild exploitation. CVSS v3.1 has the Temporal Exploit Code Maturity factor. In both, someone has to set the value and keep it current.

EPSS uses exploit evidence as a model feature. The How It Works page states that the presence of public exploit code is one of the strongest signals in the model, and that EPSS asks what kind of exploit code, published where and how widely. EPSS does not publish the feature weights, but it states exploit code and offensive tooling as a feature category among approximately 2,850 features.

## What happens on listing day

The KEV Lag Ledger lets this be checked on one set. The 318 KEV additions were confirmed by CISA as exploited, and 260 had an EPSS value on the day of listing. Of those, 197 (75.8 percent) were below 0.1, with a median of 0.0066. Many of them therefore reached KEV without a public-exploit signal strong enough to lift the score, or before that signal had accumulated.

The EPSS How It Works page explains the likely reason. Newly disclosed vulnerabilities have sparse signals, and EPSS trains on exploitation that data partners can observe. Among the 144 additions listed within 7 days, 86 had an EPSS value that day, and 81 of those 86 (94.2 percent) were below 0.1.

The conclusion for ranking is that absence of public exploit code on the day of listing is not evidence of low risk. The ledger can only count what the FIRST API recorded, but it shows that public signals trail confirmed exploitation for a sizeable share of new KEV entries.

## Malware and ransomware evidence

KEV carries a knownRansomwareCampaignUse field with values Known and Unknown. In the ledger window, 29 of 318 additions are flagged Known (9.1 percent). The median EPSS on listing day for those flagged entries was 0.0114, higher than the all-entries median of 0.0066 but still low in absolute terms.

The flag reflects the evidence CISA holds at the time of the export, so a dated export matters. Qualys states that VMDR prioritizes vulnerabilities exploited by ransomware, which is a vendor feature built on this kind of evidence; the method is not published in the page read.

## Using it in a ranking

Rank in this order: confirmed exploitation (KEV, with the ransomware flag), then assessed active exploitation (Vulnrichment), then a weaponized module, then a public proof of concept, then probability from EPSS. Each step down is weaker evidence about the present and stronger evidence about the possible.

Record the date you checked. Exploit-DB and Metasploit are archives that grow, so a CVE with no public code on one date can have a module the next. A policy that cites public exploit evidence should state which sources it checks and how often.

For a team that checks exploit code manually, the minimum record is the CVE, the source checked, the URL of the exploit or module, the date found and the date checked. For a team that relies on a platform's exploit flag, ask which of the sources above feed it.

## Limits

Public archives hold what was published. Exploitation by actors who do not publish is invisible to them, which FIRST names as observation bias in the EPSS training labels. A missing exploit is not a safe vulnerability.

This page describes the evidence types and cites the ledger for listing-day behavior. It does not count Exploit-DB or Metasploit entries per CVE, and no ledger figure here is a count of those archives.

A second limit is time. Evidence about a CVE grows after publication: an exploit appears, a module is written, a threat actor adopts it. A single check at publication is a snapshot. The 17.3 percent of KEV additions listed more than a year after publication are the clearest case, because the evidence arrived long after the record was created.

## Comparisons with this system

- [CVSS vs EPSS vs KEV](/compare/cvss-vs-epss-vs-kev/)

## Sources

- Exploit-DB, About the Exploit Database (OffSec): https://www.exploit-db.com/about-exploit-db (read 10 Nov 2026)
- Metasploit documentation home: https://docs.metasploit.com/ (read 10 Nov 2026)
- CISA, Vulnrichment README: https://github.com/cisagov/vulnrichment (read 10 Nov 2026)
- FIRST, How EPSS works: https://www.first.org/epss/how-it-works (read 10 Nov 2026)
- FIRST, CVSS v4.0 specification (Exploit Maturity): https://www.first.org/cvss/v4-0/specification-document (read 10 Nov 2026)
- CISA, KEV catalog JSON feed: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 10 Nov 2026)

Last reviewed 10 Nov 2026. Canonical: https://epssvscvss.com/scores/exploit-evidence/
