Skip to content

Likelihood of Exploitation Populations

Overview

Our ability to remediate depends on

  1. the priority (risk) of CVEs - the ones we want to remediate based on our security posture
    1. In the Understanding Risk chapter, we saw the ordered Likelihood of Exploitation for different populations of CVEs in the Risk Remediation Taxonomy.
  2. the number of CVEs for that priority (risk) - that we have the capacity/resources to fix

This section gives a

  1. view of the sizes of those populations
  2. the data sources for those populations

Population Sizes

Exploitation Venn Diagram

Population Sizes associated with the Risk Remediation Taxonomy - Likelihood of Exploitation branch.
Representative sizes and overlaps shown as there isn't authoritative exact data.
  1. ~~50% (~~100K) of all CVEs (~200K) have known exploits Proof Of Concepts available (based on a commercial CTI product used by the author)
  2. ~~5% (~10K) of all CVEs are actively exploited
    1. There isn't a single complete authoritative source for these CVEs
  3. ~~0.5% (~1K) of all CVEs (~200K) are in CISA Known Exploited Vulnerability

CVEs represent a subset of all vulnerabilities. Your organization will have a subset of these CVEs

  1. Not all exploits are public/known.
  2. Not all public exploits have CVEs.
  3. A typical enterprise will have a subset of exploits/CVEs: ~~10K order of magnitude unique CVE IDs.
    1. The counts of these unique CVE IDs may follow a Pareto type distribution i.e. there will be many instances of a small number of CVE IDs.

Likelihood of Exploitation Data Sources

This table shows the number of CVEs (from all published CVEs) that are listed in that data source:

Data Source Detail ~~ CVE count K
CISA KEV Active Exploitation 1
VulnCheck KEV Active Exploitation 2
Metasploit modules Weaponized Exploit 3
Nuclei templates Weaponized Exploit 2
ExploitDB Published Exploit Code 25

Note

VulnCheck KEV was launched just before this guide was released. So it has not been included in any analysis for this guide initial release - but will likely be for future releases.

EPSS Scores are available for all published CVEs - and cover the range of Likelihood of Exploitation from 0 to 100%.

The population sizes for Likelihood of Exploitation decrease, as Likelihood of Exploitation increases

The population sizes for higher Likelihood of Exploitation (Active ~~5%, Weaponized ~~3%) are relatively small compared to Proof Of Concept (~~50%), and All CVEs (100%).

This table lists the main public data sources.

Other Vulnerability Data Sources

In addition, there are many more Vulnerability Data Sources:

  1. Open Source vulnerability database
    1. "This infrastructure serves as an aggregator of vulnerability databases that have adopted the OSV schema, including GitHub Security Advisories, PyPA, RustSec, and Global Security Database, and more."
      1. GitHub Advisory Database
      2. Global Security Database
      3. PyPA/PYSEC
      4. RUSTSEC
  2. Red Hat Security Advisories/RHSB
  3. Go Vulnerability Database
  4. Dell Security Advisory
  5. Qualys Vulnerability database
  6. Tenable Vulnerability database
  7. Trickest "Almost every publicly available CVE PoC"

Cyber Threat Intelligence vendors may provide an aggregation of this data.

Takeaways

  1. There isn't a single complete authoritative source for all CVEs that are actively exploited - so we need to use multiple incomplete imperfect sources.
  2. The population sizes for higher Likelihood of Exploitation (Active ~~5%, Weaponized ~~3%) are relatively small compared to Proof Of Concept (~~50%), and All CVEs (100%).
  3. Not all vulnerabilities are public/known, and for those that are known, not all of them have CVEs assigned.
  4. A typical enterprise will have a subset of exploits/CVEs: ~~10K order of magnitude unique CVE IDs.
    1. The counts of these unique CVE IDs may follow a Pareto type distribution i.e. for your environment, there will likely be many instances of a small number of CVE IDs.