Application Security

Authentication

The customer portal supports email-and-password authentication with JSON Web Tokens stored in secure HTTP-only cookies. This limits direct JavaScript access to session tokens and reduces exposure to common token theft through cross-site scripting.

Enterprise SSO can connect to approved identity providers through SAML or OAuth-based flows.

Session Management

• —Strong, rotated signing secrets
• —Secure, HTTP-only, same-site cookies
• —Appropriate session expiry
• —Reauthentication for high-risk actions
• —Logout and revocation support
• —Device and anomaly monitoring where required

Role-Based Authorization

Rumbe separates customer, agent, tenant-admin, and platform-admin capabilities. Protected routes and server operations verify both the role and the active organization.

Least privilege should be applied to knowledge management, exports, provider keys, billing, user administration, and audit logs.

API Validation

Zod or equivalent schemas can validate API inputs before business logic runs. This reduces malformed data, type confusion, and unexpected payload structures.

Validation should be combined with authorization, parameterized database access, size limits, file validation, rate limiting, and safe error handling.

Safe AI Output Rendering

Rumbe renders Markdown and code blocks through a controlled parser rather than inserting raw model output directly into the page. The rendering layer should sanitize links and HTML to reduce injection risk.

Webhook Security

Stripe and other inbound webhooks should be cryptographically verified before processing. Replay handling, idempotency, event logging, and least-privilege endpoint access are important for billing and provisioning integrity.

Workflow Traceability

Chat transfers, ticket status changes, agent state, exports, and administrative changes can be recorded so investigators can reconstruct high-impact actions.

Security Testing

Application testing should cover authentication bypass, authorization boundaries, tenant manipulation, injection, file upload, session fixation, cross-site scripting, cross-site request forgery, rate abuse, webhook replay, and insecure direct object references.