Vulnerability prioritization best practices

In short: first enable the catalogs (CISA KEV, CERT-FR ALE), then adjust the CVSS-BTE and EPSS thresholds, and finally adapt each asset category to its confidentiality, integrity and availability (CIA) requirements and to its exposure.

This guide offers an approach and starting values to configure your prioritization policies. It complements the page Use the prioritization policy to prioritize vulnerabilities, which describes in detail how the policies work.

Why prioritize

Tens of thousands of CVEs are published every year, far more than a team can fix. Without a prioritization policy, everything seems urgent and the truly dangerous vulnerabilities get lost in the mass.

On a mixed fleet of workstations and servers, here is the effect of each method on the number of CVEs to process:

MethodCriteriaUnique CVEsShare of total
No prioritizationAll the CVEs present11,825100%
One dimensionCVSS-BTE ≥ 81,95717%
Two dimensionsCVSS-BTE ≥ 8 and EPSS ≥ 1%5755%
CatalogsPresence in a catalog (exploited CVEs)1992%

The catalogs isolate the 2% of actively exploited CVEs, to be fixed as an absolute priority.

Understand the CVSS-BTE score

The CVSS-BTE score is the CVSS v4 score of a vulnerability, recalculated with the context of the threat and that of your asset. The “raw” CVSS score published with a CVE describes the vulnerability in absolute terms; the CVSS-BTE indicates its actual severity on a given asset, at a given time. If a CVE has no CVSS v4 score, Cyberwatch applies the same principle to its CVSS v3 score, or failing that CVSS v2.

The three metric groups

The acronym follows the nomenclature of the CVSS v4.0 standard published by FIRST: each letter matches a metric group taken into account in the calculation.

LetterMetric groupWhat it describesSource in Cyberwatch
BBaseIntrinsic characteristics of the vulnerability: attack vector, complexity, privileges required, impactScore published with the CVE
TThreatExploit maturity: vulnerability under attack, public proof of concept or no known exploitExploit level detected by Cyberwatch, see Vulnerabilities score
EEnvironmentalImportance of the asset (CIA requirements) and exploitation conditions specific to your environmentCIA requirements and CVSS ceilings of the asset’s prioritization policy

The combinations give four types of scores:

  • CVSS-B: base metrics only;
  • CVSS-BT: base and threat;
  • CVSS-BE: base and environmental;
  • CVSS-BTE: base, threat and environmental.

The type used by a policy depends on the metrics enabled in it.

How context changes the score

The CVSS standard treats a metric that is not filled in as the worst case. The base score therefore assumes that an exploit is used and that the asset is highly sensitive. The threat and environmental metrics only refine this assumption: in practice, the CVSS-BTE is equal to or lower than the base score.

Example with a vulnerability that can be exploited remotely without authentication, with a high impact (vector 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):

ContextAdded metricsTypeScore
No context—CVSS-B9.3
“Functional” or “High” exploitE:ACVSS-BT9.3
“Proof of Concept” exploitE:PCVSS-BT8.9
“Unavailable” exploitE:UCVSS-BT8.1
Isolated asset, attack vector capped at “Local”MAV:LCVSS-BE8.6
Low-sensitivity asset (low CIA requirements)CR:L/IR:L/AR:LCVSS-BE8.9
No known exploit on a low-sensitivity assetE:U/CR:L/IR:L/AR:LCVSS-BTE6.5

With a threshold of 7, this CVE is prioritized on a sensitive server, but no longer on a low-sensitivity asset as long as no exploit is known. If an exploit appears, Cyberwatch recalculates the score and the CVE becomes prioritized again.

Available criteria

A CVE is prioritized on an asset if it appears in a catalog selected by the policy, or if it reaches all the configured thresholds that apply to it (at least one must apply). An empty or zero threshold is ignored, as is a criterion for which the CVE has no score.

CriterionWhat it measuresLogicCaveat
Catalogs (CISA KEV, CERT-FR ALE, BSI…)Observed active exploitationSufficient on its ownFoundation of any policy
EPSS (0 to 100%)Probability of exploitation within 30 daysCombined with the other thresholdsChanges daily
Contextualized CVSS score (CVSS-B, CVSS-BT, CVSS-BE or CVSS-BTE)Severity adjusted to the threat and to the assetCombined with the other thresholdsSee Understand the CVSS-BTE score
SSVC decisionHandling urgency according to the CISA methodCombined with the other thresholdsOnly CVEs processed by CISA have SSVC metrics; the SSVC section of the policy must be filled in
Exploit levelMaturity of the exploit codeCombined with the other thresholdsAvailable since version 15.10; evaluated only if the maturity is known

CVSS ceilings complement these criteria: they cap the base metrics for a given environment. For example, an attack vector capped at “Local” for an asset disconnected from the network lowers the score of remotely exploitable CVEs.

Prioritization is built in layers: first what is exploited, then what is severe and likely, and finally what matters for the business of the asset.

  1. Enable the catalogs on all policies. At a minimum CISA KEV and CERT-FR ALE. These CVEs are actively exploited and should be fixed immediately, whatever their score.
  2. Classify assets by criticality and exposure. Distinguish at least production and pre-production, DMZ and internal network, servers and workstations. The same CVE does not have the same urgency on an exposed server and on an internal workstation.
  3. Use the CVSS-BTE score. Enable the threat and environmental metrics so that the score reflects the actual exploitability and the CIA requirements of the asset.
  4. Add an EPSS threshold to the CVSS threshold. In the example above, the CVSS-BTE and EPSS pair divides the volume by three compared to CVSS alone (575 CVEs versus 1,957).
  5. Tighten for exposed assets, relax for the others. High CIA requirements and lower thresholds in the DMZ; a more lenient policy for the internal network and pre-production.
  6. Refine if needed. CVSS ceilings for isolated assets, SSVC criterion, exploit level criterion.
  7. Measure and readjust. Check that the volume of high-priority CVEs remains manageable for your teams; otherwise raise the thresholds, without ever removing the catalogs.

Typical policies by asset category

Recommended starting point: one policy per criticality level, with the catalogs enabled everywhere. The three default policies cover most cases. The “Medium” policy is applied to new assets. The default policies cannot be edited: any variant requires a custom policy.

Asset categoryPolicyCatalogsCVSS-BTE thresholdEPSS thresholdCIA requirements
Exposed to the Internet, DMZ, core networkCustom “Critical”All6.00.1%High
Sensitive business productionHigh (default)All7.00.2%High
Standard internal serversMedium (default)All7.00.5%Medium
WorkstationsCustomAll9.0As neededMedium
Pre-production, development, isolated assetsLow (default)All——Low

Thresholds are minimum values: a CVE whose score is equal to the threshold reaches it. The thresholds of the “Low” policy (CVSS 10 and EPSS 100%) in practice only retain the CVEs from the catalogs. For an asset disconnected from the network, add a CVSS ceiling with a “Local” attack vector. These values are starting points: adjust them according to the volume your teams can actually fix and to your regulatory obligations.

Assign and maintain the policies

Assign the policies through assets rules rather than manually: new assets then automatically receive the right policy.

  • Assets rules. Filter the inventory (hostname, group, operating system…), then use “Bulk edits > Create a rule” with the “Set prioritization policy” action. Example: any asset whose hostname contains dev receives the Development group and the “Low” policy. See Rules.
  • One-off bulk edit. Use “Inventory > Bulk edits > Update the prioritization policy” for a batch of existing assets.
  • Network exposure (since version 16.0). Filter the inventory by network exposure to isolate the assets exposed to the Internet and the most heavily accessed ones, then create a rule that assigns them the strictest policy.
  • Periodic review. Check that no critical asset has remained on the default policy, especially after a discovery or a bulk import.

Drive remediation

A policy is only valuable if it drives action: associate each policy with a remediation deadline and track overruns.

  • Response time. Define a maximum time between detection and fix for each policy: the default policies offer 30 days (High), 60 days (Medium) and 180 days (Low). Overdue vulnerabilities receive a red badge showing the number of overdue days in the “Vulnerabilities” and “Patch management” tabs. The “Summary” tab of the asset compares the average remediation time with the objective.
  • Patches first. Process first the patches at the top of the “Patch management” tab and of the “Corrective actions” page: they are sorted by number of high-priority vulnerabilities fixed.
  • Alerts. In the vulnerability encyclopedia, filter on “Status: Prioritized”, “Catalog: All” and “Severity: Critical”, then use “Bulk edits > Create an alert” to be notified of new CVEs matching these criteria.
  • Reports. Use the executive summary: when the artificial intelligence features are enabled, it highlights the main risks of the fleet and up to four prioritized recommendations.

Common pitfalls and limitations

Pitfall or limitationConsequenceGood practice
Removing the catalogs to reduce the volumeExploited CVEs drop out of the prioritiesRaise the CVSS or EPSS thresholds and keep the catalogs
Prioritizing on the CVSS score alone17% of CVEs remain prioritized in the exampleCombine CVSS-BTE and EPSS
Leaving all assets on “Medium”No difference between the DMZ and the internal networkCreate assets rules by criticality and exposure
Relying on SSVC for all CVEsOnly CVEs processed by CISA have an SSVC decisionUse SSVC as a complement, never alone
Wanting several sets of thresholds on the same asset (for example CVSS ≥ 7 and EPSS ≥ 80%, or CVSS ≥ 8 and EPSS ≥ 50%)Impossible: an asset has a single policy, whose thresholds are combinedChoose one pair of thresholds per asset category
Confusing exploit level with presence in CISA KEVWaiting for a “High” exploit for a CVE already in the KEV catalogThe KEV catalog is taken into account by the catalog criterion, regardless of the exploit level
Editing a default policyImpossibleCreate a custom policy

Going further