githubEdit

lockSecurity and Secret Encryption

Flashback applies a consistent secret-encryption model across the platform. This model is not limited to Cloud Storage: it also applies to AI resources and to any connected resource that requires credentials, tokens, API keys, or private material.

Why this is a dedicated platform topic

Secrets are used in multiple places, including:

  • bucket and container credentials,

  • AI provider credentials,

  • repository-level integrations,

  • and future resource connectors added to the gateway.

Because the same trust model is reused everywhere, encryption and decryption behavior should be understood as a platform security principle rather than a feature of one page or one connector.

Core model: encrypted at rest, isolated decryption path

All secrets in Flashback are encrypted in the backend database using an RSA-based mechanism where the backend service itself does not hold the complete ability to decrypt plaintext secrets directly.

This gives two practical outcomes:

  1. Secrets remain unreadable as plaintext in persistent backend storage.

  2. Secret decryption is moved away from the regular backend request path.

triangle-exclamation

Decryption lifecycle on Bridge Nodes

Decryption is handled by Bridge Nodes through a federated shared-secret process:

  • On startup, a Bridge Node retrieves encrypted secret payloads needed for authorized operations.

  • Decryption occurs only inside the node runtime in memory.

  • Plaintext secrets are not written to disk or long-term storage.

  • When the node stops, in-memory secret material is purged and must be re-established on next startup.

This reduces persistent exposure and avoids concentrating decryption capability in a single always-on backend surface.

Scope of application

The same model applies whenever Flashback needs to act on behalf of your configured resources, including:

  • Cloud Storage connectors,

  • AI/LLM connectors,

  • and other gateway resources that use delegated credentials.

As new resource types are introduced, they inherit this same security baseline so governance and operational behavior stay consistent across the platform.

Repository API key secret handling

Repository API keys (storage and AI) follow the same platform model and add important operational constraints:

  • API key secrets are encrypted at rest in backend storage.

  • Full plaintext secret values are shown only once in the UI at key-creation time.

  • After the creation view is closed/reloaded, the secret cannot be retrieved again.

  • If the secret is lost, you must generate a new key and rotate client credentials.

triangle-exclamation

For repository operations and key-management workflows, see:

Operational guidance for teams

To use this model effectively in production:

  • Keep original credentials in your own enterprise secret-management process.

  • Rotate provider keys and tokens based on your security policy.

  • Use least-privilege scopes when creating external credentials.

  • Review Bridge Node operational security (runtime hardening, access control, and monitoring).

Last updated

Was this helpful?