Offloading Cryptography with the Vault Transit Engine

The Problem: Key Sprawl and Exposure

When applications handle encryption, keys end up in environment variables, config files, or application memory. The core risk is key sprawl, where the secret used to encrypt data exists across dozens of instances.

The Vault Transit Engine treats cryptography as a service. The application sends plaintext to Vault and receives ciphertext back. The plaintext key never leaves Vault's secure boundary.

How Transit Operates

Transit provides API endpoints for symmetric and asymmetric cryptographic operations. Supported algorithms include AES-GCM as the default for symmetric encryption, ChaCha20Poly1305, and RSA, ECDSA, Ed25519 for asymmetric needs.

Key versioning is central. Rotating a key creates a new version instead of overwriting the old one. Vault tracks which version encrypted a value, allowing seamless decryption of older data while new encryption uses the latest version.

Symmetric Encryption Workflow

The following assumes Vault 1.12+ and administrative permissions to configure the engine.

Enable the Transit secret engine on your local terminal with an authenticated Vault session.

# Enable the transit engine at the 'transit' path
vault secrets enable transit

Create a named key. Vault generates the key material internally.

# Create a key with the default AES-GCM algorithm
vault write -f transit/keys/app-payment-key

Encrypt data. Vault expects plaintext via the CLI.

# Encrypt the string 'secret-credit-card-number'
vault transit encrypt app-payment-key secret-credit-card-number

Vault returns a ciphertext string starting with vault:v1:. The v1 indicates the key version used.

Decrypt data by sending the ciphertext back.

# Decrypt the ciphertext returned from the previous step
vault transit decrypt app-payment-key vault:v1:encrypted-data-string

Performance Trade-offs and Limitations

Offloading crypto improves security but adds a network round-trip per operation.

• Latency: Every crypto operation is an API call. High throughput workloads increase response times.
• CPU Saturation: Cryptographic operations are CPU intensive. Size Vault nodes for expected request volume.
• Policy Risk: Overly broad policies granting update on transit/decrypt to many roles allow unintended decryption.

Verifying the Implementation

Confirm the mount is active.

vault transit list-endpoints

Perform a round-trip test with a known plaintext and verify the decrypted output matches exactly.

Inspect audit logs for encrypt and decrypt events. Key material should never appear in logs.

To rotate a key, use vault write -f transit/keys/app-payment-key/rotate. New data uses the new version while old data remains decryptable.