Security & architecture

All data in transit is protected with TLS 1.3 by default. Data at rest is encrypted using industry-standard encryption at the storage layer. Self-hosted deployments can configure their own encryption keys (BYOK).

Sensitive configuration values are encrypted at the application layer before being written to the database. The encryption key is configured via a server-side secret in your environment.

Authentication tokens are short-lived JWTs (1-hour access tokens) signed with HS256 or ES256. Refresh tokens are stored as opaque, single-use values.

The PrivateFlow cloud service runs on hardened infrastructure with the following controls:

• Network isolation with private VPCs and no direct database internet exposure
• Row-Level Security (RLS) enforced at the database layer - queries cannot cross organizational boundaries even if application code is compromised
• Rate limiting on all API endpoints (configurable per deployment)
• SSE endpoints limited to 10 concurrent connections per user to prevent resource exhaustion
• Dependency updates via automated PR scanning (Renovate)
• Container images are rebuilt weekly with the latest base image patches

We follow OWASP Top 10 mitigations across the codebase:

• Injection: All database queries use parameterized statements. Route handlers validate all input with strict schema validation - no unsafe casts.
• Broken access control: Multi-tenant row isolation is enforced structurally on every query. Admin routes require elevated role claims in the authenticated session.
• SSRF: All externally supplied URLs are validated against an allowlist before forwarding requests.
• Env var injection: Environment variable references in flow configurations are validated against an approved allowlist - arbitrary variable access is blocked.
• XSS: Generative UI component trees are sanitized to strip dangerous event handlers and unsafe HTML injection vectors.
• Prompt injection: A dedicated prompt injection guardrail scans incoming user messages for adversarial patterns before they reach the LLM.

Every user action that modifies data (create/update/delete flows, runs, API keys, team members, SSO connections) is written to an audit log designed to be tamper-evident, with actor, IP, resource type, and outcome. Audit logs are accessible to org admins via the Compliance dashboard.

OpenTelemetry tracing and metrics can be enabled in your deployment configuration, with export to any OTLP-compatible backend.

For hybrid deployments, only audit metadata is synced to the central hub by design - no PII, no prompt content, no model outputs. The sync is opt-in per node and disabled by default.

We operate a responsible disclosure program. If you discover a security vulnerability in PrivateFlow (cloud service or self-hosted engine), please report it privately before public disclosure.

How to report: Email [email protected] with:

• A description of the vulnerability and its potential impact
• Steps to reproduce (proof-of-concept if available)
• Any affected versions or deployment modes

We will acknowledge your report within 48 hours and aim to release a fix within 90 days for critical vulnerabilities. We do not currently offer monetary bounties, but we will credit researchers in our security advisories unless they prefer to remain anonymous.

Please do not: exploit the vulnerability beyond what is needed to demonstrate it, access or modify data that does not belong to you, or perform denial-of-service testing against our production infrastructure.

For non-vulnerability security questions (compliance documentation, penetration test authorization, enterprise security reviews):

[email protected]

For enterprise customers, dedicated security review sessions and custom DPAs (Data Processing Agreements) are available. Contact [email protected].