Home / Trust / Information Security Policy
Security policy

Information Security Policy

How HealthOS protects the information it holds, who is accountable, how information is classified and handled, and the suppliers involved.

Purpose

HealthOS holds health information about people on behalf of the organisations that care for them, and connects to government systems that trust it to behave. This policy states how that information is protected, who is accountable, and which documents carry the detail. Every other policy in this directory sits under this one.

Scope

  • Information: customer data (personal and health information held for a tenant), information obtained from government systems, source code, credentials and keys, and the organisation's own business records.
  • Systems: every deployed HealthOS stack on Azure, the GitHub repositories and CI, the marketing websites, and the devices and accounts used to build and operate them.
  • People: everyone covered by Personnel Security.

Customer-hosted installations are in scope for the software and out of scope for the hosting, which is the customer's.

Objectives

  1. Confidentiality. A person's information is seen only by people with a reason to see it, and one tenant's information never reaches another.
  2. Integrity. Records are accurate, and every change is attributable.
  3. Availability. The service and its data can be restored within the times in infra/BACKUP-AND-RECOVERY.md.
  4. Compliance. The Privacy Act 1988 and the Australian Privacy Principles, the My Health Records Act and Healthcare Identifiers Act, the NDIS Act where NDIA information is held, and the terms of each government and licensor agreement are met.

How well these are being met is measured by these measures, reported to the director quarterly in the ISMS review: open findings by severity; the age of the oldest unapplied critical or high patch; the last restore drill date and result; incidents and near misses; access reviews completed on time; training completion.

Roles and accountability

RoleWhoAccountable for
DirectorsAaron Le Saux (sole director)Approving this policy and the risk appetite; resourcing security
Security ownerAaron Le SauxThis policy set, the risk register, the incident response, assessor liaison
Privacy officerAaron Le Saux (privacy@healthos.net.au)Privacy Act compliance, data breach assessment and notification
EngineeringEveryone who changes the systemBuilding to the rules below, and reporting what they see
Everyone—Following this policy and reporting suspected incidents at once

Principles

These are how the system is built, and each can be checked in the repository.

  • Tenant isolation in three layers. Every tenant route requires an organisation context, runs in an organisation-scoped transaction, and reads tables under forced row-level security. The runtime database role cannot bypass it. A check (verify:tenant-isolation) fails a deployment where that is not true.
  • Least privilege. Role-based access with facility and ward scope inside the application; a separate, narrower database role for the patient portal; managed identities instead of stored keys where the platform allows.
  • Secure by default. Production refuses to start on a missing, short or reused signing secret. Mock and sandbox modes are refused in production unless explicitly allowed. Outbound connections to private networks and plain HTTP are off in production.
  • Defence in depth. No single control is relied on: a route check and row-level security; platform encryption and field encryption for credentials; a classifier and separate signing secrets per token class.
  • Everything is attributable. Reads and changes of clinical records are logged in append-only tables kept seven years (Audit Logging).
  • Data stays in Australia. Stacks are deployed to Azure australiaeast. The one qualification is recorded under suppliers below.

Information classification and handling

ClassWhat it isHandling
RestrictedHealth information and government identifiers (IHI, Medicare, DVA, NDIS numbers); information obtained from government systems, including NDIA Protected Agency Information; credentials, keys and certificates; licensed classification materialStored only in the production data stores and vaults, encrypted at rest and in transit. Never in a repository, a ticket, an email, a chat, a log or a lower environment. Access is by role and is logged. Licensed classification material has further rules in the repository's CLAUDE.md
ConfidentialOther personal information; customer configuration and contracts; source code; security assessments, pen test reports and this risk register; unreleased plansShared on a need-to-know basis inside the organisation. Sent outside only under a confidentiality agreement and over an encrypted channel
InternalWorking documents, procedures, internal communicationsNot published; no special handling beyond that
PublicThe website, published guides, published specificationsNo restriction

Rules that follow from the table:

  • No production data outside production. Development and testing use the seeded fictional tenants, never a copy of a customer's data (secure-development.md).
  • Support access to customer data happens only on a customer's request, through a named account, and is logged like any other access.
  • Retention. Clinical and audit records are kept for the periods the law and the customer's own obligations set; SaaS workspace retention is in saas-data-retention.md.
  • Disposal. Data is deleted from the live stores on the retention schedule and ages out of backups within the backup retention (35 days). Equipment that held Confidential or Restricted information is wiped before reuse or disposal.
  • De-identified is not public. De-identified clinical text sent to an AI service is still handled as Confidential.

Access control

  • Every account belongs to one named person. Shared accounts are not used; where a platform forces a shared administrator credential it is held in Key Vault and its use is logged.
  • Multi-factor authentication is required on GitHub, Azure and every application administrator account.
  • Passwords are checked against known breaches and stored as bcrypt hashes; the product's own password and session policy is set per organisation.
  • Access is granted on request and approval, revalidated every six months, and removed on the day someone leaves or changes role (backup-restoration-and-privileged-access.md).

Acceptable use

People covered by this policy:

  • use organisation systems and customer data only for the work they are engaged to do;
  • work on devices that have full-disk encryption, a screen lock, a supported and patched operating system, and are not shared;
  • keep credentials in a password manager, never in a file, a note or a message, and never reuse one across services;
  • do not copy Restricted information to a personal device, personal cloud storage or removable media;
  • do not paste Restricted information into any external tool or AI service that is not on the supplier list below for that purpose;
  • report a lost device, a suspicious message or a mistake at once. Reporting a mistake quickly is the expected behaviour and is never penalised.

The device standard: full-disk encryption on; screen lock at 5 minutes or less; a supported operating system with automatic updates; built-in anti-malware on; no shared accounts. The owner checks it quarterly and records the check in SharePoint: HealthOS ISMS > Asset and device register.

Suppliers

A supplier that stores or processes Confidential or Restricted information is assessed before use and reviewed every year. The assessment records what the supplier holds, where, under what terms, and what independent assurance it has (a SOC 2 report or ISO 27001 certificate, read, not just noted).

SupplierWhat it doesHoldsWhere
Microsoft AzureHosting, database, storage, Key Vault, monitoringRestrictedaustraliaeast
Azure AI FoundryAI model for coding featuresConfidential (de-identified clinical text)Account in australiaeast. No Restricted information is sent to it
GitHubSource hosting, CIConfidential (source code)Outside Australia
AI coding toolsDevelopment assistanceConfidential (source code)Outside Australia
StripeSubscription billingSubscriber billing detailsOutside Australia
Microsoft 365HealthOS's own email (sign-up links, invitations, sign-in codes, support), sent app-only with Mail.Send from no-reply@healthos.net.au in HealthOS's own tenancy. Postmark was used before and is decommissionedEmail addresses, message contentMicrosoft's data centres for HealthOS's Microsoft 365 tenancy
TelstraTelephony and SMSPhone numbers, message contentAustralia
Kelly PartnersAccountingHealthOS's financial records, subscriber billing detailsAustralia
PentiSecurity scanning of the deployed applicationWhat an unauthenticated or test-account scan can see, and the findingsIn Penti's platform in the United States (an overseas disclosure; the results describe vulnerabilities, not personal information)

This table is drawn from the repository and is not complete. The full list, with the contract or terms each supplier is used under and the date its assurance report was last read, is kept in SharePoint: HealthOS ISMS > Supplier register. [Owner to set: the domain registrar and any support tooling still to be added.]

No supplier is given Restricted information unless it is hosted in Australia or the customer and, where relevant, the government agency concerned has agreed.

The policy set

SubjectDocument
RiskRisk Management Framework
Incidents and data breachesSecurity Incident and Data Breach Response
PeoplePersonnel Security
Encryption and keysEncryption and Key Management, key-rotation.md
Audit loggingAudit Logging
Development and changesecure-development.md
Patching and vulnerabilitiespatching-and-vulnerability-management.md
Customer notificationend-of-support-and-vulnerability-notification.md
Backup, restore, privileged accessbackup-restoration-and-privileged-access.md, infra/BACKUP-AND-RECOVERY.md
Security testingpentest/README.md
Vulnerability disclosurewebsite/.well-known/security.txt and the trust page

Exceptions

An exception to any policy in this set is requested in writing with its reason, the risk it creates, what compensates for it, and an end date. It is approved by the security owner, or by a director where the security owner is the requester, and entered in the risk register. An exception without an end date is not an exception; it is a change to the policy and is made as one.

Breaches of this policy

A breach is reported like any other incident. A deliberate or repeated breach may lead to removal of access and to action under the person's agreement.

Questions about security? Email security@healthos.net.au