Skip to main content

Security

Security and data handling

See where Nebula processes information, how access works and what evidence is kept.

Data flow

How customer material moves through Nebula.

Each step names the service, purpose and safeguard. The public white paper uses the same view.

Protected path

Five steps, with a safeguard at each boundary.

Access, processing, storage, AI and proof are controlled separately.

  1. 01

    Customer boundary

    People and source material

    People use authenticated sessions to reach organisation and engagement workspaces.

  2. 02

    Nebula application

    Application access controls

    Server routes apply session, organisation, membership and route-specific access checks.

  3. 03

    Managed data services

    Operational records and files

    The providers' current documentation states that their managed database and object-storage services encrypt stored data at rest. Nebula has not retained current account or applicable-contract evidence establishing the deployed boundary, key custody or backup coverage.

  4. 04

    Configured AI providers

    Document reading

    Redaction applies to configured text fields by default, not necessarily filenames or native PDFs. Provider terms require confirmation.

  5. 05

    Proof boundary

    Confirmation records

    Eligible events can produce a Polygon hash and a signed Nebulad payload. Source documents are excluded.

Access
Authenticated sessions and organisation or engagement membership gate the relevant workspace routes.
Storage
Structured records and stored files remain separate managed services with application-controlled access paths.
Reading
Available providers use the same selection rules and Nebula data policy. Deployed redaction settings and provider terms require confirmation.
Proof
Each confirmation uses its matching canonical preimage; payloads are not interchangeable.
Customer access, Nebula processing, managed storage, configured AI paths and eligible proof output.

Review paths

Start with the question you need answered.

01Data handlingWhere information goes, how it is protected and what follows.

01 · Data architecture

Function configuration and managed data services

Source configuration specifies Sydney for Vercel Functions; deployed region unverified. Records use managed Neon Postgres and documents use Cloudflare R2.

Function instruction
Vercel / syd1 in source
Database
Neon / region unverified
Object storage
Cloudflare R2 / location unverified

02 · Encryption at rest

Encryption at rest

The providers' current documentation states that their managed database and object-storage services encrypt stored data at rest. Nebula has not retained current account or applicable-contract evidence establishing the deployed boundary, key custody or backup coverage. Separately, Nebula wraps each user's signing key with AES-256-GCM using a runtime key. Production sessions use signed, encrypted, httpOnly and secure cookies.

Database
Provider-managed; account evidence pending
Object storage
Provider-managed; account evidence pending
Signing keys
Wrapped by a runtime-only KEK
Sessions
Signed and encrypted cookies

03 · Encryption in transit

TLS and strict transport security

Application responses set HSTS for two years, including subdomains, and named provider endpoints use HTTPS. Deployed database and provider settings, including negotiated TLS versions, still require verification.

Application transport
HTTPS and HSTS configured
Provider transport
Deployed settings require verification
HSTS
Two years, includeSubDomains

04 · Privacy posture

AI data use and text redaction

Nebula removes common identifiers from configured prompt and extracted-text fields before provider processing by default. This covers phone numbers, emails, TFNs, Medicare numbers and driver licence numbers. Public ABNs and ACNs remain. Filenames and native PDF files are not guaranteed to be redacted, and deployment settings can disable this control.

After a 30-day grace period, erasure anonymises the account and selected name copies, disables sign-in, revokes keys and removes notifications. Shared engagement, audit and proof records may remain. Retained canonical proof data can include identifiers and event details.

Model training
Nebula policy excludes customer data
Text redaction
Default for configured text fields; deployment configurable
Erasure
30-day grace, then anonymised
02Identity and accessSign-in, membership, elevated access and account controls.

05 · Who can see what

Engagement access and support access

Signed-in areas require a session. Engagement areas also require party membership. Machine endpoints use their own authentication. Super-admin access passes a separate gate. When Nebula staff enter an organisation's work through support access, Nebula records the acting identity, stated reason, organisation, each engagement reached, and the start and end. The affected organisation can review that read-only record in its account.

Password accounts support TOTP multi-factor authentication with single-use recovery codes, and an organisation can require MFA for its members.

Signed-in API routes
Shared session gate
Engagement routes
Membership checked
Public and machine routes
Route-specific authentication
Nebula staff support access
Separate gate; record visible to the affected organisation
Counterparty access
By invitation, revocable

06 · Account protection

Rate-limited sign-in and multi-factor authentication

Password sign-in is rate-limited. Users can enable TOTP multi-factor authentication with single-use recovery codes, and an organisation can require MFA for password accounts.

Password hashing
bcrypt / cost 12
Sign-in limit
10 attempts per 15 minutes per client identifier
MFA
TOTP and single-use recovery codes
Organisation policy
MFA can be required for password accounts
03Service operationsActivity records and visible handling requirements.

07 · Audit trail

Access and activity logs

After an API request passes or fails the shared authentication gate, Nebula attempts to record the caller, route, method, decision, user agent and salted IP hash. A 90-day pruning path is configured, but its production execution has not been verified. Recorded engagement changes and selected exports have separate audit paths.

Access log
Best-effort shared auth-gate write
Retention
90-day pruning path configured
IP addresses
Salted hash in the AccessLog row

08 · Classified material

Classification labels for handling context

Nebula records and displays four Australian Government classification labels: UNCLASSIFIED, OFFICIAL, OFFICIAL: SENSITIVE and PROTECTED. Labels keep handling requirements visible. Access follows assigned roles, while personnel clearances and handling authority remain part of the customer's approved process.

Before material above UNCLASSIFIED enters an environment, Nebula maps the applicable handling requirements and authority with the customer. A higher-classification deployment follows customer-specific control selection, independent assessment and formal authorisation before use.

Labels
Four Australian Government labels
Default
UNCLASSIFIED
Higher classifications
Controls selected, assessed and authorised for the deployment
04AI and assuranceModel safeguards and the standards shaping Nebula's controls.

09 · Document reading

Shared safeguards across available providers

Each reading uses an available provider and the model selected for that session. Nebula removes common identifiers from configured prompt and extracted-text fields by default, but filenames and native PDF files are not guaranteed to be redacted. Nebula's policy excludes customer content from model training. Provider terms and the deployed redaction setting must be confirmed for each customer scope.

Model selection
Available model selected for the session
Configured text fields
Identifier redaction enabled by default
Filenames and native PDFs
Not guaranteed to be redacted
Model training
Excluded by Nebula policy; provider terms pending

10 · Frameworks and posture

Control design mapped to SOC 2 criteria and informed by the ISM

SOC 2 is an independent CPA examination and report on controls relevant to security, availability, processing integrity, confidentiality and privacy. Nebula maps its control design and evidence to the AICPA Trust Services Criteria used for that examination.

The Australian Government Information Security Manual is the Australian Signals Directorate's risk-based framework for protecting systems and data. Nebula uses it to organise its control and data-handling matrix. IRAP is the assessment pathway for an ASD-endorsed assessor to review an agreed system boundary against applicable ISM controls. Where procurement requires that evidence, the customer agrees the deployment scope, assessor and authorisation decision.

SOC 2
Trust Services Criteria shape the control and evidence mapping
ISM
Australian Government risk-based guidance informs the control matrix
IRAP
Deployment-specific assessment pathway where procurement requires it

The white paper

Read the Nebula white paper.

The public paper covers authority, architecture, service regions, AI data flows, proof, standards and procurement review.

Receive the public paper by email

Use the form for an emailed copy. The cover opens the same paper.

Submit the form to receive the white paper at your work email.

Common questions

Security, answered plainly.

First questions, answered from these boundaries.

Where is Nebula's data processed and stored?

Source code specifies Sydney for Vercel Functions; deployment unverified. The providers' current documentation states that their managed database and object-storage services encrypt stored data at rest. Nebula has not retained current account or applicable-contract evidence establishing the deployed boundary, key custody or backup coverage. Reading providers may process data outside Australia; terms determine location.

Is Nebula SOC 2 certified or IRAP assessed?

No certification is claimed. Controls are mapped to the SOC 2 Trust Services Criteria and informed by the Australian Government ISM. IRAP scope, assessor and authorisation are agreed with the customer.

Is customer data used for advertising or shared model training?

Nebula's product policy excludes customer documents, their contents and private derived information from model training and cross customer patterns. The signed in application carries no advertising or analytics tracking, and public pages use consent gated measurement only. Provider terms require confirmation for the customer scope.

Can we review Nebula's security before buying?

Yes. The white paper and architecture view are public; deployment-specific evidence is agreed during Enterprise review.

How does Nebula help prove a record has not been altered?

Key engagement moments can be anchored by hash. An anchor attests integrity and timing, not the claim's truth; disputed records remain visibly disputed.

How do we report a security issue?

Email security@nebulaplatform.com.au. The draft vulnerability disclosure policy sets the proposed scope and handling process.