Vulnerability Management Program: Scanning, Prioritization and SLAs

In what order should vulnerabilities be closed? Credentialed scanning, what CVSS, EPSS and CISA KEV each tell you, asset criticality, SLAs, the exception process and audit evidence.
Technical summary
Technology
CVSS v4.0, EPSS, CISA KEV, SSVC, vulnerability scanners
Topic
Designing an end-to-end vulnerability management program
Level
Intermediate
Prerequisites
  • An up-to-date asset inventory
  • Patch deployment infrastructure (WSUS, Intune, etc.)
In this article8 sections

A scanner that runs once a month, a report hundreds of pages long, a critical finding count that stays the same every month. In many organizations, that’s all vulnerability management is, and in an audit the question “how do you manage vulnerabilities?” is answered with a screenshot of a scan. Producing reports isn’t a program. Below we build, step by step, a vulnerability management program that works through scope, credentialed scanning, prioritization by exploitation signals, remediation deadlines, an exception process and measurable metrics.

The difference between vulnerability scanning and vulnerability management

Scanning collects data; management turns this data into decisions and closure. From an enterprise cybersecurity standpoint, the components of a program are:

  1. Scope: Which assets are assessed, with which method and how often.
  2. Detection: Network-based scanning, agent-based scanning, cloud configuration assessment, application scanning.
  3. Prioritization: A combination of severity, exploitability and asset criticality.
  4. Remediation: A patch, a configuration change, a version upgrade or a compensating control.
  5. Verification: Confirming the fix with a rescan.
  6. Exceptions and risk acceptance: Time-limited, approved management of findings that can’t be closed.
  7. Measurement: Metrics showing the program is working.

ISO/IEC 27001:2022 Annex A 8.8 (Management of technical vulnerabilities) and NIS2’s vulnerability handling area want to see not that scanning is done, but that this cycle turns. For the technical side of patch deployment (WSUS, Azure Update Manager), see the server patch management guide; here we focus on the program’s decision and measurement layer.

Scope and credentialed scanning

Scope is determined by the asset inventory, not by the devices the scanner “finds”. Report two lists separately: assets in the inventory that haven’t been scanned in the last 30 days, and assets that show up in scans but aren’t in the inventory. The program’s blind spots are in these two lists.

A network scan without credentials only assesses externally visible services and banner information. Installed software versions, missing operating system patches and local configuration errors only show up in credentialed scanning or with an agent.

Method Strength Limitation
Uncredentialed network scan Shows the attacker’s view Misses local software vulnerabilities
Credentialed network scan Sees patches and software versions Requires privileged account management
Agent-based scan Also covers laptops outside the network Can’t be installed on network devices and embedded systems
External surface scan Confirms internet-facing assets Doesn’t show risk inside the internal network

The credentialed scanning account is a valuable target for an attacker. On Windows, this account should be local administrator only on scan targets, shouldn’t be able to sign in interactively and shouldn’t be a domain admin. On Linux, it should work with an SSH key and limited sudo rights. Track the share of targets where authentication succeeded in every scan. A common picture: a scanning account whose password has changed, hundreds of targets where authentication failed, and a report that looks “clean”. A clean report misleads. For discovering internet-facing assets, the external attack surface management guide is a complement.

The limits of CVSS: prioritizing with EPSS and KEV

The CVSS base score measures a vulnerability’s technical severity; it measures neither the risk in your environment nor the likelihood of exploitation. Programs that treat only CVSS 7.0 and above as “urgent” put thousands of findings at the same priority and can’t tell apart the ones really being exploited. When the team sees everything as urgent, it can’t do anything urgently. That’s also why CVSS v4.0 clearly emphasizes enriching the base score with threat and environmental metrics.

Two external signals are useful in practice for prioritization:

  • The CISA KEV catalog: The catalog in which US CISA lists vulnerabilities with reliable evidence of active exploitation in the wild. Being on the list means “real, not theoretical”.
  • EPSS: A model published by FIRST that estimates, on a 0–1 scale, the probability that a CVE will be seen in exploitation activity within the next 30 days. It’s updated daily.

Both sources are offered in machine-readable form and can be added to scan results automatically:

# Download the CISA KEV catalog and extract the CVE list
curl -s https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json \
  | jq -r '.vulnerabilities[].cveID' > kev-cve.txt

# EPSS probability and percentile for a single CVE
curl -s "https://api.first.org/data/v1/epss?cve=CVE-2021-44228" \
  | jq '.data[] | {cve, epss, percentile}'

The decision logic combines these signals with asset criticality. The ranking below is an example template; adjust the thresholds to your own risk appetite:

  1. Findings in KEV on internet-facing or critical assets: highest priority.
  2. Internal network findings in KEV, and internet-facing findings with a high EPSS value (example threshold: above 0.1).
  3. High/critical CVSS findings on critical assets with no exploitation signal.
  4. Other findings: closed in the standard maintenance cycle.

If you want a more structured decision tree, look at CISA’s SSVC (Stakeholder-Specific Vulnerability Categorization) approach. SSVC uses exploitation status, automatability of the attack, technical impact and mission/public impact as decision points, and gives the result as action categories such as “track”, “attend” and “act” rather than a numeric score.

Designing remediation deadlines (SLAs)

The SLA is the maximum time allowed from the detection date of a finding to its verified closure. Keep these three points in mind when designing it:

  • Tie the deadline to the priority class above, not to the CVSS class.
  • The clock starts at the first detection of the finding and stops at the rescan date. Closing the ticket alone isn’t evidence.
  • Take the reality of maintenance windows into account. An SLA that can’t be met turns into a pile of exceptions.
Priority class Example SLA Notes
P1: KEV + exposed/critical asset 7 days or less The emergency change process is used
P2: High exploitation signal 15 days The next maintenance window
P3: High severity, no signal 30 days Monthly patch cycle
P4: Other 90 days With the version upgrade plan

The deadlines in the table are examples; they aren’t taken from any standard. If your customer contract or sector regulation specifies a shorter deadline, that deadline applies.

The exception and risk acceptance process

Some findings don’t close within the SLA: an application whose vendor has ended support, a certified medical device, a production server that isn’t allowed to restart. For OT systems that can’t be patched, the decision method is explained separately in the OT vulnerability management and patch decisions guide. In the IT environment, an exception record should include at least:

  • The finding, affected assets and priority class
  • The reason it can’t be closed (technical obstacle, vendor dependency)
  • Compensating controls: network isolation, disabling the service, access restriction, additional monitoring
  • Approval from the business unit’s manager as the risk owner
  • An end date; open-ended exceptions aren’t accepted

An expired exception automatically turns back into an open finding and is reported as an SLA breach. On renewal, verify that the compensating control is still in place; the firewall rule may have quietly disappeared a few changes later. A rule on paper provides no protection.

Program metrics

Metrics should answer not “how many findings are there?” but “is the program working?”:

Metric What it shows
Scan coverage (scanned / assets in inventory) Size of the blind spot
Authentication success rate Reliability of the results
Average time to close by priority class Remediation speed
Share of findings closed within the SLA Commitment compliance
Number and age of open findings past the SLA Accumulated risk
Number of reopened findings Non-lasting fixes
Number of active and expired exceptions Discipline of risk acceptance

When you report metrics by asset-owning team, responsibility for remediation becomes clear. If most reopened findings are caused by configuration, the problem isn’t in patching but in the baseline; this topic is covered in the configuration compliance and drift tracking guide.

An evidence set for audits

The auditor questions the program’s design and its operation separately. The set below is accepted as sufficient in most ISO 27001, TISAX, customer and cybersecurity audits:

  • An approved vulnerability management procedure: scope, frequency, priority classes and SLA table
  • The scan schedule and dated outputs of recent scan jobs
  • A coverage report: comparison of the inventory with scanned assets
  • A sample: for a few selected critical findings, the detection date, ticket, change record and evidence of closure with a rescan
  • The current exception list, approvals and end dates
  • The metric report for the latest period and a record showing it was presented to management

Chained attack paths and business logic flaws that automated scanning can’t find require manual testing; we explain the distinction in the penetration test scope and finding verification guide. If you want to build the program from scratch or have the coverage and quality of your current scans measured, take a look at our vulnerability analysis and security scanning service. Sources: the CISA KEV catalog and FIRST EPSS.

Frequently asked questions

Is every vulnerability with a high CVSS score urgent?

No. The CVSS base score measures technical severity; it doesn’t measure the likelihood of exploitation or the impact in your environment. When presence in the CISA KEV catalog, EPSS probability and asset criticality are considered together, the really urgent findings stand out.

Why is credentialed scanning necessary?

Uncredentialed scanning only assesses services visible from the network. Missing operating system patches, installed software versions and local configuration errors only show up with credentialed scanning or an agent. Track the authentication success rate separately too.

Is there a standard value for vulnerability remediation deadlines?

There’s no universal value. The deadlines are set according to the organization’s risk appetite, maintenance windows, and contractual or regulatory requirements. Tying them to priority classes that include the exploitation signal and asset criticality, rather than to the CVSS class, is recommended.

How do you manage a vulnerability that can’t be closed?

An exception record is opened with the justification, compensating controls, the risk owner’s approval and an end date. An expired exception automatically turns back into an open finding; on renewal, it’s verified that the compensating controls are still in place.

Support

Contact us for support or a discovery session on this topic.

Contact form
[email protected]

Eretek Cybersecurity

Firewall, VPN, endpoint security and threat prevention. We reduce the attack surface and measure and report the risks.

Knowledge Center

Other articles