Files
git.stella-ops.org/docs/technical/architecture/security-boundaries.md
2025-12-24 12:38:14 +02:00

46 lines
2.3 KiB
Markdown

# Security boundaries (platform-level)
This document describes the baseline security boundaries expected across StellaOps modules. Module dossiers may impose stricter requirements.
## Authentication and authorization
All externally reachable services are expected to enforce:
1. Token validation (short-lived, tenant-scoped access tokens).
2. Sender constraints where configured (DPoP / mTLS).
3. Scope-based authorization (least privilege).
4. Tenant isolation: requests and data access are filtered by tenant context.
### Hard gates (typical examples)
Exact gates are module-specific, but common patterns include:
- **Authority**: nonce-based sender constraints (DPoP), strict token lifetimes, tenant-scoped issuance, and rate limiting.
- **Signing/attestation services**: narrow scopes, service identity requirements (often mTLS), and verification of the artifact being signed/attested (for example digest checks) before producing evidence.
Authoritative references:
- `docs/security/scopes-and-roles.md`
- `docs/modules/authority/architecture.md`
- `docs/modules/signer/architecture.md`
- `docs/modules/attestor/architecture.md`
## Network segmentation (typical deployment)
- **Front door / ingress**: TLS termination, rate limiting, and WAF controls.
- **Gateway layer**: authenticated routing and tenant resolution; exposes only required service surfaces.
- **Private service network**: internal service-to-service traffic (least privilege, explicit allowlists).
- **Stateful infrastructure**: PostgreSQL, Valkey, object storage, message brokers (not directly internet-exposed).
Deployment bundles under `deploy/` are the authoritative source of concrete network layouts.
## Data protection
- TLS for in-transit protection (including internal traffic where required by the profile).
- Secrets must not be embedded in documentation examples; use a secrets provider (file, Docker secrets, Kubernetes secrets, vault).
- Evidence artifacts should be written to content-addressed or immutable stores where replay/audit requires it.
## Auditability
The platform is designed for audits:
- Deterministic outputs for the same inputs.
- Evidence artifacts (SBOM slices, advisories/VEX observations, explain traces) linkable to digests.
- Structured logs and correlation IDs for tracing request paths across services.