Assurance policies
You already know which findings matter: a known exploited vulnerability in an image you ship, a secret in a repository, a critical bug nobody has touched in a month. An assurance policy writes that down once, so Dwaar can tell you every resource that breaks it, today and after every scan.
Dwaar evaluates assurance policies itself, from what it already holds: findings, scans, settings posture, discovery and vendor reviews. Nothing has to be installed for them to work.
What a policy is made of
| Part | What it decides |
|---|---|
| Resource type | Code, Artifact, Cloud or Vendor. A policy judges one type. |
| Organization | The organization that owns the policy. It judges resources of that organization and the ones below it. |
| Scope | Which of those resources: organizations, products, conditions and bypass entries. |
| Controls | What the policy evaluates on each resource in scope, each control with its parameters, severity and what it does on a failure. |
| Mode | Observe, Audit or Enforce. See Guard Center. |
Every save creates a new revision. Each verdict and event names the revision it was judged under, so you can always tell which version of a policy said what.
The controls
A control fails or passes on one resource, or, for the settings posture controls, on one connection. Findings controls count open findings only: not fixed, not ignored and not accepted, on the versions Dwaar tracks now. Accepting a finding's risk takes it out of every control.
| Control | Types | Fails when |
|---|---|---|
| No open vulnerabilities at or above a severity | Code, Artifact | An open finding of the chosen types is at or above the severity. You can count only findings with a fix, or only findings older than a number of days. |
| No known exploited vulnerabilities | Code, Artifact, Vendor | An open finding's vulnerability is in the CISA KEV catalog. |
| No likely exploited vulnerabilities (EPSS) | Code, Artifact | An open finding's EPSS score is at or above the minimum. |
| Fixes applied within the SLA | Code, Artifact, Vendor | An open finding with a fix is older than its severity allows (by default 7, 30, 90 and 180 days). |
| No secrets | Code, Artifact | There is an open secret finding. |
| No malicious packages | Code, Artifact | There is an open malicious package finding. |
| No weak cryptography | Code, Artifact | There is an open crypto finding at or above the severity. |
| CI workflows are safe | Code | There is an open CI workflow finding at or above the severity. |
| Base image is current | Artifact | There is an open base image finding at or above the severity. |
| Allowed licenses only | Code, Artifact | There is an open license finding at or above the severity, or a component declares a license on the denied list (exact SPDX ids). |
| No banned components | Code, Artifact | A component matches one of your package patterns, optionally with a version. |
| SBOM present | Code, Artifact | A tracked version's last full scan produced no SBOM. |
| Scanned recently | Code, Artifact | The last full scan failed, or finished longer ago than the limit. |
| Required scanners ran | Code, Artifact | A tracked version's last full scan did not cover a required scanner category. |
| Owned by a product | Code, Artifact, Cloud, Vendor | No product uses the resource, and optionally its lineage is not mapped. |
| Source control settings pass | Code | The GitHub organization has an open settings posture finding at or above the severity. |
| Default branches are protected, Two-factor authentication required | Code | The settings posture scan finds the default branch unprotected, or two-factor sign-in not required. |
| Cloud account posture passes | Cloud | The account has an open posture finding at or above the severity. |
| No unexpected public exposure | Cloud | A workload is reachable in a way you did not allow. |
| Risky apps are reviewed | Vendor | An app holds high-risk scopes or admin consent, and its vendor has no completed assessment. |
| Connections are healthy | Code, Artifact, Cloud, Vendor | A connection is in error, was never tested, or was not tested recently. |
| Vendors are assessed | Vendor | A vendor of the chosen tiers has no assessment completed in time. |
| Vendor documents on file | Vendor | A required document, such as a SOC 2 report, is missing, unreadable or expired. |
The Controls tab lists every control with its parameters. A control marked Needs a guard runs at a checkpoint only a guard can reach, so it cannot be added to a policy yet.
Scope: which resources a policy covers
A resource is in scope when all of these hold:
- It is of the policy's type and in the policy's organization or one below it.
- It is in one of the scope's organizations and products, when the scope names any.
- Every condition matches. A condition is an attribute, is any of or is none of, and values that may use
*. - No bypass entry matches. Each bypass entry exempts a resource on its own.
| Type | Attributes |
|---|---|
| Code | Tracked ref, repository visibility, repository name |
| Artifact | Registry, image repository, tracked ref |
| Cloud | Workload kind, region, exposure |
| Vendor | Kind (vendor or SaaS app), tier |
Settings posture findings belong to a connection, which has no products or attributes. A policy whose scope names products or conditions does not cover connections.
When Dwaar evaluates
You never press a button. A resource is evaluated about 30 seconds after something it depends on changes: a scan finishes, its findings change, a settings posture scan or discovery completes, a risk is accepted, or the resource joins a product. Editing a policy evaluates everything in its scope again.
Controls that depend on time, such as a fix SLA or a document's end date, are evaluated again when they can change, and every resource in scope is evaluated once a day.
Verdicts and compliance
Each policy gives each resource in scope one verdict.
| Verdict | Meaning |
|---|---|
| Compliant | Every enabled control passes. |
| Non-compliant | At least one control fails. |
| Not evaluated | Dwaar does not have what the policy needs yet, such as a first full scan. This is never counted as compliant. |
Compliance counts each resource once, over the policies in audit and enforce. A resource is non-compliant when any of those policies finds it non-compliant and no exception covers the failure. It is compliant when every one of them finds it compliant. The compliance rate is compliant resources divided by compliant plus non-compliant ones. Resources that are not evaluated are shown beside it, never in it. Compliance is its own figure: it does not change a resource's grade or its finding counts.
Guard Center, then Events, shows six tiles for your scope: compliant resources with the rate, non-compliant resources, not evaluated resources, recent events (opened or reopened in the last 24 hours), open events and overdue events.
Default policies
Every tenant starts with five policies, all in Observe: Code baseline, Source control posture, Artifact baseline, Cloud baseline and Vendor baseline. They show what would fail before anyone is told. You can change them or archive them, but not delete them.
Try this next
Open the Code baseline on Policies, read its compliance, and look at the resources it finds non-compliant. If the list looks right, move it to Audit and watch the events it opens.