Security Threat Model (Baseline)¶
Method¶
STRIDE-aligned baseline for public-facing GPUaaS control plane.
Assets¶
- User identity and access tokens
- Node access credentials and SSH materials
- Billing ledger and payment events
- Allocation and usage records
- Admin operations and audit trails
Key Threats and Mitigations¶
Spoofing¶
- Threat: token theft/session replay
- Controls: short-lived tokens, rotation, device/session anomaly checks
Tampering¶
- Threat: modification of billing/ledger records
- Controls: immutable ledger entries, transactional writes, audit logs
Repudiation¶
- Threat: admin action denial
- Controls: signed/immutable audit logs with actor and correlation IDs
Information Disclosure¶
- Threat: secret leakage, over-broad API response
- Controls: secret manager, least-privilege data access, response minimization
Denial of Service¶
- Threat: API abuse/websocket floods
- Controls: WAF, rate limits, backpressure, autoscaling thresholds
Elevation of Privilege¶
- Threat: bypassing tenant/role checks
- Controls: centralized authorization policy enforcement + service-side checks
Priority Threat Areas (MVP)¶
- Payment/webhook replay and tampering
- Provisioning privilege abuse
- Multi-tenant data boundary leakage
- Secret exfiltration via logs/config