Trust
Security at Bella
Bella holds salon operators’ guest data, payroll data, and payment data. We treat that responsibility seriously. This page describes the controls we have in place — technical, organizational, and contractual.
Encryption
- In transit: TLS 1.2 or higher is enforced on every endpoint. Older protocols (TLS 1.0, TLS 1.1, SSL) are disabled. HSTS is set with a long max-age so browsers refuse plaintext downgrades.
- At rest: AES-256 on all primary storage and backups. Database disks, object storage, and snapshots are all encrypted with keys managed by GCP KMS.
- Secrets: Application secrets (API keys, webhook signing secrets, third-party tokens) are stored in a configuration vault, never in source code, never in logs.
Tenant isolation
Bella is multi-tenant by design. Tenant-owned tables carry a tenant key (company_id), and every authenticated request resolves to a tenant scope through request-scope middleware before any database call. On financial-write paths (audit log, staging queue), database-trigger invariants additionally verify that the row's tenant + realm + connection identity match a registered connection — cross-tenant writes fail closed at the database layer. Continued route-by-route auditing of scope enforcement is on the SOC 2 readiness program.
Access control
- Role-based access control (Owner, Manager, Stylist) at the application layer. Each role has explicit, audited capabilities.
- Internal staff access to customer data requires a documented support ticket and is logged. Access is revoked immediately on separation.
- Multi-factor authentication required for all internal admin accounts and recommended for customer accounts.
- Production database access is restricted to a small named set of engineers via SSH bastion + key authentication.
Network & perimeter
- Cloudflare in front of every public endpoint: DDoS protection, WAF, rate limiting.
- VPC-level firewall rules: only ports 80/443 open to the public internet, everything else (database, internal services, MCP servers) is private.
- API gateway enforces request signing on Stripe webhooks (signature verification rejects any request with an invalid signature). Twilio, Telnyx, and Plaid webhook signature verification is on the security roadmap; in the interim, these endpoints rely on payload-hash replay defence and IP-source attribution.
Monitoring & incident response
- Append-only audit and provenance logs of every state-changing action: who, what, when, from where. Retained for 7 years per the Data Retention & Disposal Policy.
- Real-time alerting on suspicious patterns (auth failures, rate-limit hits, signature failures, unusual data export volume).
- Incident response runbook with named on-call. Customer notification within 72 hours of confirmed breach — faster if material.
- Backups: daily, encrypted at rest with provider-managed keys. Restore testing is a documented SOC 2 readiness commitment performed at least annually.
Compliance
- SOC 2: readiness program active; Type I readiness target within 12 months, followed by a Type II observation period after readiness gaps are remediated. Interim policy documentation and control evidence available under NDA.
- PCI DSS: reduced scope. Card processing is delegated to Stripe; Bella does not store or process full card numbers. Bella's responsibilities are limited to SAQ-A-equivalent controls.
- TCPA / CASL: built into the SMS engine — consent is enforced at the database layer, opt-out is honored within seconds, every message has a downloadable certificate of consent.
- HIPAA: we do not knowingly process PHI. Don’t put medical data in your guest notes.
- GDPR / CCPA: data subject rights are honored within 30 days; see Privacy Policy.
Vulnerability disclosure
If you believe you’ve found a security issue, email [email protected]. We commit to:
- Acknowledging your report within 24 hours.
- Investigating in good faith and keeping you informed.
- Not pursuing legal action against good-faith researchers who follow standard responsible-disclosure practices (no destructive testing, no data exfiltration beyond proof of concept, give us reasonable time to fix).
- Crediting researchers in our public security acknowledgments page (with your permission).
Sub-processors
See the Subprocessors page for the authoritative current list, with purpose, data categories, and region per subprocessor. Each subprocessor is reviewed for security posture before onboarding and re-reviewed annually.
Questions? Contact us or request a demo.