Security Model

Core principles

• Client-side encryption for sensitive credential values.
• Passphrase never transmitted as plaintext to backend services.
• Database stores encrypted blobs and vault metadata.
• In-memory vault keying with lock/unlock lifecycle.

Vault formats

New format (preferred)

• vault_config.raw_dek: base64-encoded data encryption key material.
• vault_config.bcrypt_hash: passphrase verifier.
• Unlock sequence: bcrypt verify passphrase -> import DEK -> enable decrypt operations.

Legacy format (compatibility)

• vault_config.wrapped_dek remains supported.
• Unlock sequence derives KEK from passphrase and unwraps DEK.

Locking

• Manual and automatic locking supported.
• Auto-lock defaults to 15 minutes of inactivity.
• Locking removes active key references from runtime state.

Browser-side protections

• CSP is configured through ContentSecurityPolicy.ts with environment-dependent policy variants.
• Security event logging tracks unlock failures, decrypt errors, and suspicious patterns.

Threat boundaries

Keyper protects strongly against database disclosure and passive network interception of stored secrets. Like all browser apps, it still depends on endpoint/browser integrity during active sessions.

Multi-user boundaries

• Each username maps to an independent vault context (vault_config.user_id) with separate key material.
• Self-service registration creates new vault metadata without introducing privileged/admin access paths.
• User switching changes context only; it does not grant access until that user’s passphrase is verified.
• Emergency passphrase reset remains self-service and scoped to the selected user’s bcrypt_hash record.