Skip to content
SkyRosterBook a discovery call

Trust · Security

Security controls and the evidence to review

Review authentication, permissions, document protection and the configuration of your deployment together. These mechanisms describe the product architecture; current test evidence and operational settings belong in your security assessment.

1 · Authentication

Your identity provider decides who you are, not us

SkyRoster uses a separate Keycloak realm for each tenant and signed OAuth2/OIDC tokens for authenticated API access. The gateway verifies token signatures using an identity provider's signing keys. The trusted issuer configuration, token lifetime, audience and identity-provider policies must be checked for the environment being deployed. Standard single sign-on is part of Core; tenant setup and any custom identity work are confirmed in the written offer.

verified

Your staff do not need a second password. SkyRoster federates login against the identity provider you already run:

  • Microsoft Entra ID (Azure AD), through Keycloak's built-in Microsoft OIDC identity provider. An optional "autologin" mode skips SkyRoster's own login screen entirely and redirects straight to your Microsoft sign-in, while administrative accounts can be kept on a standard username-and-password screen for break-glass access if your identity provider is ever unreachable.
  • On-premises Active Directory, through Keycloak's LDAP federation, with optional Kerberos for a fully passwordless login from inside your own network. A separate, one-directional sync job can also read your Active Directory to create matching employee records automatically as people join, though it does not remove someone from SkyRoster when they leave your directory; that step stays with whoever administers your organisation here.
configured

2 · The internal call gate

Keep service access within its intended boundary

The intended deployment exposes the gateway and keeps business services on internal networks. The gateway removes the caller-supplied internal-call marker before routing a request. User-facing calls carry the user's security context; scheduled jobs and message consumers also have controlled service contexts. Network exposure, service privileges and internal-only routes must be checked in the deployed environment.

verified

3 · Authorisation

Permissions at the action and data boundaries

The product uses several authorisation mechanisms. Their coverage needs to be verified for each workflow, including direct API calls:

  1. Endpoint permissions. Protected controllers check the resource and membership tier required for an operation.
  2. Command and query permissions. Protected handlers can enforce the same operation permissions when a call reaches them through another application path.
  3. Record scope. Secured repositories apply employee and unit restrictions. Other persistence paths need explicit checks for the records an operation can access.

The full model behind what "the membership tier the action requires" means, the sixteen areas it applies to and how read and write are scoped independently, is on the access control page.

4 · The secured repository layer

Test record access through the API

Secured repositories can add the caller's permitted units and employees to a query before it executes and validate writes against the permitted record scope. Tenant-specific databases provide a separate boundary between organisations. Tenant isolation and employee-level access are different controls: both belong in acceptance tests for the workflows and roles your organisation uses.

verified

5 · Transport

Encrypted in transit, and the gateway is the only door

The deployment model uses HTTPS for public application traffic and a gateway cross-origin policy for permitted browser origins. Certificate renewal, exposed endpoints, internal transport and operational-tool access must be checked against your chosen hosting setup. A source-code description alone does not establish those running-environment controls.

verified

6 · Stored documents

Encrypted before they are written, decrypted only to be read

Employees attach documents under five categories: personal, managerial, qualification, leave and duty. That covers IDs, licences, medical certificates and qualification proofs. File content is encrypted at the point of storage and decrypted only at the moment it is read; this has been true for every document uploaded since the product's 4.11.0 release, and a small number of files uploaded before that version are recognised and read through a separate, backward-compatible path rather than silently treated as encrypted when they are not.

General employee responses separate confidential personal fields from ordinary roster data. Dedicated confidential-data endpoints apply additional permissions, with a separate self-service route for an employee's own information. Audit records and exports have their own access paths and may retain historical personal values; include those paths when reviewing permissions, retention and the roles allowed to inspect history.

verified

Security review

Security testing that leaves a regression test behind

Ask for the assessment scope, tested product version, outstanding findings and remediation evidence relevant to your deployment. Regression tests should cover authenticated roles, record ownership, cross-tenant access and historical data as well as normal user journeys. This page does not assert that every security finding is closed or that source inspection proves the security of a running environment.

Read next

The rest of the trust section

Bring us a month that went wrong.

Start with a 20-minute discovery call to discuss your operation and priorities. We can then arrange a tailored working session with your roster and rules.

Book a discovery call

Booking opens a 20-minute discovery call. A tailored working session is arranged afterwards.

Send a question instead