Skip to main content
Version: 0.4 · current

Security posture

ConceptReviewed October 11, 2026
Applies to:Excubius 0.4Roles: Viewer, Operator, Editor, Entity admin
PlansPro · EnterpriseDeploymentSelf-hosted

The security modules look for weaknesses in how your databases, identities and cloud resources are configured, and report each one as a finding with its evidence, the baseline it is measured against and a suggested action. Findings feed a security score from 0 to 100 per area and overall, with its history. This is detection and posture only: Excubius never changes a monitored system, and the suggested action is text for a person to follow in a change window.

From checks to a score​

The security modules are part of the Pro plan as an add-on and of the Enterprise plan. A super administrator switches them on for the installation; an administrator then adds the checks to resources like any other catalog check.

Findings and alerts​

A finding is an alert of the category Security findings, so it has everything an alert has: the inbox, seen-by, assignment, the timeline and Jira. The Security posture page adds the area, the evidence, the baseline reference, the suggested action and the risk acceptances. What makes a finding different is that it describes posture, not health.

AlertFinding
DescribesSomething is failing or out of its limits nowA configuration that weakens your security
Resource stateTurns the resource amber or redNever changes the resource's state
Counts inThe open alertsThe open alerts, and the security score of its area
NotifiedThrough the entity's channelsThrough the entity's channels; recommendations never are
Can be accepted as a riskNoYes, by an editor, with a reason and an end date
Resolves whenThe check recovers, or by handThe next successful run no longer detects it, or by hand for events

Some findings are events rather than states: a change in the members of sysadmin, of privileged directory groups, or of Owner and Contributor role assignments in Azure. An event is reviewed, not fixed: resolve it by hand once someone has checked it. Events cannot be accepted as a risk.

Recommendations are the medium and low assessments of Defender for Cloud. They are listed separately and count in the score only while the installation shows them. High-severity assessments are findings like any other.

The life of a finding​

Accepting a risk​

Sometimes a finding is known and tolerated for a while: a legacy application that needs a setting until it is replaced, for example. An Editor or higher can then accept the risk with Accept risk…:

  • The Reason explains why the risk is acceptable and what mitigates it, in 3 to 2000 characters.
  • The acceptance ends on a date between one hour and one year ahead.
  • While it lasts, the finding leaves the open list, shows Accepted risk and does not count in the score. Excubius keeps checking it.
  • The finding opens again on its own when the acceptance ends, or as soon as its severity rises. An editor can also Reopen it earlier.
  • Every acceptance is kept with who accepted it, when, until when and how it ended: expired, reopened, replaced by a new acceptance, the severity rose, or resolved.

The security score​

The score is calculated for every entity and everything below it, in three areas: Identity, Database and Cloud. Findings of other modules, such as host hardening, are listed under Other and not scored.

  1. Each open finding takes points from the score of its resource: critical 25, warning 10, info 2. A finding weighs up to 50% more the longer it stays open, reaching the maximum after 90 days.
  2. A resource scores 100 minus its points, never below 0. A resource counts in an area once one of the area's checks has run on it, or once it has a finding there.
  3. An area scores the average of its resources, weighted by their criticality, from 1 to 5, and 3 when not set. An area with no assessed resource shows Not assessed yet and is left out of the overall score.
  4. The overall score is the average of the assessed areas, weighted by the area weights of the entity. Each weight is 1 unless an entity admin changes it; weights are inherited down the tree.

For example, sql-ventas-01 has criticality 5, sa enabled for 30 days (25 × 1.17 = 29.2 points) and one new warning (10 points), so it scores 60.8. sql-pruebas-01 has criticality 1 and no findings, so it scores 100. The database area of their entity scores (5 × 60.8 + 1 × 100) / 6 = 67.4.

Accepted risks do not count while the acceptance lasts. How the score is calculated on the Security posture page repeats these rules next to the numbers.

History. Excubius keeps a snapshot of every entity's scores each hour. After 14 days the snapshots are reduced to one a day and kept for 400 days. The page shows the last 30 days and the change over 7 days.

Where findings come from​

SourceAreaWhat it brings
Native SQL Server checksDatabaseThe sa login, command execution features, surface area, linked-server credentials, privileged role changes, failed logins, server audit, backup encryption and orphaned users
Microsoft Entra ID and Active DirectoryIdentityPrivileged roles without MFA, risky sign-ins, changes to privileged groups, stale accounts
Azure Resource GraphCloudManagement ports open to the Internet, public storage, Key Vault and backup vault protection, Owner and Contributor changes
Microsoft Defender for CloudCloudIts unhealthy assessments, one finding per assessment and resource, with Defender's remediation text as the suggested action
WazuhCloudAlerts of level 7 and above from the last 24 hours, and critical and high vulnerabilities, matched to the host of each agent

Defender for Cloud and Wazuh are optional sources, each switched on separately. Vendor findings keep one finding per item across runs, and an item the vendor stops reporting resolves. Switching a source off stops collecting from it; the findings it already raised stay open until someone resolves them.

Every source is read with read-only permissions: Reader and Security Reader in Azure, read-only Microsoft Graph application permissions, a read-only SQL Server login and a read-only Wazuh user. The Wazuh client sends only searches. Wazuh agents are deployed by your own tooling, never by Excubius.