Document Version: 1.1
Last Updated: 2025-10-03
Owners: Platform Security Guild
This checklist summarises how each Lilith component addresses the six STRIDE threat categories. Status tags follow this convention:
- ✅ Implemented & continuously monitored
- 🟡 Partially implemented — tracked on the security backlog with delivery dates
- 🔴 Not started — requires immediate attention (none outstanding as of this revision)
| Component | Spoofing | Tampering | Repudiation | Information Disclosure | Denial of Service | Elevation of Privilege |
|---|---|---|---|---|---|---|
| BFF Service | ✅ mTLS between services, JWT issuer/audience validation, strict correlation IDs | ✅ Input validation layer, schema coercion, response signing for downstream calls | ✅ Structured request logging with immutable correlation IDs | ✅ Response scrubbing, pii-safe serializers, security headers | ✅ Rate limiting, circuit breakers, backpressure aware SSE streaming | ✅ Persona-level RBAC enforcement, service-to-service scopes |
| Authentication Service | ✅ WebAuthn, OAuth device binding, rotated signing keys | ✅ Encrypted JWT signing keys in HSM, signature verification | ✅ Audit ledger with user/session fingerprinting | ✅ Bcrypt+pepper, encrypted secrets, zero knowledge proof for password resets | ✅ Adaptive throttling, incremental lockouts, CAPTCHA escalation | ✅ Role lattice enforced in token claims, privileged path allow-lists |
| Conversation Service | ✅ Request signing, auth context propagation | ✅ Message digests, append-only histories, integrity checks | ✅ Event sourcing with append-only logs and tamper detection | ✅ Field-level encryption for transcripts, scoped retrieval APIs | ✅ Connection quotas, sliding window limits, fan-out protection | ✅ Persona isolation, per-conversation ACLs, administrative approvals |
| AI Orchestration | ✅ Provider certificate pinning, signed provider manifests | ✅ Prompt firewall, context sanitisation, deterministic streaming filters | ✅ Prompt/response hashing with provenance logs | ✅ Prompt redaction, fine-grained context segmentation, data minimisation | ✅ Adaptive budget caps, queue backoffs, provider failover mesh | ✅ Policy engine gating tool-call privileges, trust zone separation |
| Content Service | ✅ Signed ingestion manifests, curator authentication | ✅ Version snapshots with hash verification, immutable history | ✅ Editorial workflow ledger with reviewer signatures | ✅ Attribute-based filtering, zero-trust content access | ✅ Staged publish pipeline with cache warming guardrails | ✅ Moderator/curator RBAC, principle of least privilege |
| Media Service | ✅ Signed URL issuance, token-bound playback | ✅ Integrity checksums, tamper-evident waveform manifests | ✅ Delivery receipts with request signing | ✅ DRM encryption, least-privilege object policies | ✅ Tiered CDN with origin shield, adaptive bitrate protections | ✅ Admin-only transcoding queues, separation of duties |
| Notification Service | ✅ Mutual auth with providers, OAuth client secrets rotation | ✅ Template immutable storage, checksum validation | ✅ Delivery logs with verifiable receipts | ✅ Data minimisation in payloads, encryption at rest | ✅ Queue depth guards, exponential backoff, provider circuit breaking | ✅ Scoped API keys, tenant-aware topic ACLs |
| Data Stores (PostgreSQL, Redis, Object Storage) | ✅ IAM role separation, certificate auth, Vault-issued creds | ✅ Row level security, Crypto signatures for critical tables | ✅ WAL shipping with append-only archives, tamper warnings | ✅ Transparent data encryption, field level tokenisation | ✅ Connection pooling limits, query governors, protective indexes | ✅ DB roles scoped per service, automated rotation |
| Observability / Telemetry Stack | ✅ Authenticated ingestion endpoints, signed agent configs | ✅ Immutable log pipeline with sigstore attestations | ✅ WORM storage buckets, retained trace IDs | ✅ PII scrubbing filters, differential privacy budgets | ✅ Multi-tenant rate limiting, log ingestion throttles | ✅ RBAC dashboards, break-glass workflow with approvals |
| Client Applications (Web & Mobile) | ✅ PKCE/OAuth, signed release artefacts, app attestation | ✅ Runtime integrity checks, CSP + SRI enforcement | ✅ In-app audit journaling, analytics event signing | ✅ Secure storage APIs, privacy-safe analytics, consent gating | ✅ Offline request quotas, exponential backoff, offline cache expiration | ✅ RBAC capabilities in-app, hardened admin consoles |
Governance & Continuous Improvement#
- Review cadence: Quarterly security guild review; critical updates on demand
- Metrics tracked: Unresolved STRIDE items, time-to-mitigation, number of backlog exceptions
- Tooling: Threat Dragon for modelling snapshots, Sigstore for artefact assurance, Backstage security catalog for control owners
Any change to architecture must re-run this checklist and update artefacts
referenced by tests/security.threat-model.spec.ts.