Security best practices
A checklist for keeping keys, tokens, users and data safe in a Lynkoria integration.
Keys and secrets
- Keep service keys and webhook secrets in your backend’s secret manager. Never ship them in a web or mobile bundle, build variable, URL or log.
- Give each deployment its own key with only the scopes it needs, and name it after that deployment.
- Use
lynk_test_keys in staging andlynk_live_keys in production. - Rotate keys on a schedule and when people leave; revoke immediately if a key is exposed, then review the project’s Activity tab.
- Set an expiry on keys for temporary integrations.
Users and tokens
- Issue tokens only from an authenticated endpoint protected against CSRF.
- Derive
userId,deviceId,tenantIdand call roles on the server; never trust them from the request. - Check membership and your product rules before creating conversations, issuing join tokens, or promoting livestream speakers.
- Create a new client when a different user signs in, and disconnect the old one.
- Do not log tokens. They are short-lived, but they grant access until they expire.
Browser access
- List only your exact production origins under Allowed origins for the production environment. Keep
localhostin staging only. - Render message bodies as text. Never inject them as HTML.
Webhooks
- Verify every webhook signature on the raw body with the SDK, and reject timestamps older than 5 minutes (the SDK does this).
- Deduplicate on event
id. - Use an HTTPS endpoint on an origin your platform team has approved.
Content and data
- Attachments are not malware-scanned. Restrict allowed types in the messaging policy, and scan files yourself if users exchange untrusted documents.
- Set a message retention period if your product must not keep messages indefinitely.
- Presence is visible across a tenant; partition sensitive groups into separate tenants.
- Media is encrypted in transit (DTLS-SRTP) but is not end-to-end encrypted. Do not promise end-to-end encryption to users.
Operational controls you have
| Situation | Action |
|---|---|
| Key exposed | Revoke it in the console; rotate dependent deployments |
| Abusive user | users.suspend(userId); remove them from calls |
| Need to stop a call now | calls.end(callId) |
| Need to stop a whole environment | Disable it in the console, then end active calls |
Products handling regulated or sensitive data need their own security and privacy review before launch. The platform team’s view is in platform security.