SecurityCenter
Security Center
Trust Through Transparency
Every action is logged. Every privilege is gated. Every prediction is auditable. This is the security backbone enterprise buyers actually require.
Request Lifecycle
Every privileged action: like running an AI scan: passes through these gates.
STEP 1
Frontend Request
React app calls base44.functions.invoke() over HTTPS with auth token
STEP 2
Auth Verify
Function calls auth.me(): invalid token → 401
STEP 3
Role Gate
Admin-only paths check user.role === 'admin' → 403 if not
STEP 4
Execute + Log
Run logic with service role; write AuditLogEntry
STEP 5
Response
JSON returned over HTTPS; UI updates via React Query
Security Posture
ACTIVE
Authentication
- ✓Email + Google SSO via Base44
- ✓Token-based sessions (no passwords stored)
- ✓Every backend function verifies identity via createClientFromRequest
- ✓Auth errors → automatic redirect to login
ACTIVE
Authorization (RBAC)
- ✓Roles: admin / user (extensible)
- ✓Admin-only functions enforce user.role === "admin"
- ✓User entity: only admins can list/manage other users
- ✓Per-function authorization gates
ACTIVE
Data Isolation
- ✓Test & Production databases fully separated
- ✓No code changes required to switch environments
- ✓Test mistakes cannot corrupt production data
- ✓JSON Schema validation on every write
ACTIVE
Backend Privilege Levels
- ✓User-scoped: base44.entities (respects RLS)
- ✓Service role: base44.asServiceRole (admin tasks only)
- ✓Service role used only after auth verification
- ✓Never exposed to frontend
ACTIVE
Secrets Management
- ✓API keys in encrypted environment variables
- ✓Accessed only inside Deno functions
- ✓Never sent to the browser
- ✓Never committed to source control
ACTIVE
Transport & Storage
- ✓HTTPS enforced for all traffic
- ✓Public file storage for asset images
- ✓Private storage with signed URLs available
- ✓File URLs only: no base64 payloads in DB
ACTIVE
AI Data Handling
- ✓LLM calls proxied via Base44 (Gemini / GPT / Claude)
- ✓Provider keys never reach client
- ✓Image-based analysis uses gemini_3_1_pro
- ✓Every prediction logged to PredictionAccuracy
ACTIVE
Audit & Observability
- ✓Immutable AuditLogEntry records
- ✓Actor identity verified server-side (never trusted from client)
- ✓IP hint + user agent captured per action
- ✓Filterable, exportable audit log viewer (admin only)
ACTIVE
Verifiable Outcomes
- ✓PredictionAccuracy ledger of every AI call
- ✓Verified Savings Ledger ties predictions → dollars
- ✓Human verification required for "verified" status
- ✓Defensible audit trail for buyers & regulators
Audit log is admin-only
Sign in as an administrator to view detailed activity records.