Tink for Python HOW-TO

This document presents instructions and Python code snippets for common tasks in Tink.

Setup instructions

Tink requires Python 3.7 or above.

The simplest way is to install a binary release from PyPi:

pip3 install tink

Currently, the following set of binary wheels are published:

  • Linux
    • Python 3.7
    • Python 3.8
    • Python 3.9
  • macOS
    • Python 3.7
    • Python 3.8
    • Python 3.9

In addition to the binary wheels, a source distribution is also published. If pip does not find a suitable binary wheel for your environment, it will fall back to building the project using the source distribution. This process has the same requirements as building from source instructions below.

Common setup issues

  • If you get an error regarding the root certs when using the GCP KMS integration, you will have to install the root certs.

Building from source

Tink currently supports two build systems for use with Python:

  • Bazel
  • Setuptools to create a Python package

Note that in both cases Bazel is required, as it is needed to compile the wrapper around the C++ implementation which uses pybind11.

Build with Bazel

To build the Python implementation:

cd python
bazel build ...

Build a Python package using pip

A setup script is provided which allows building a Python package using pip.

The setup script requires:

To build and install the Python package:

cd python
pip3 install .

Running tests

To run all tests, you can:

cd python
bazel test ...

Initializing Tink

Tink provides customizable initialization, which allows for choosing specific implementations (identified by key types) of desired primitives. This initialization happens via registration of the implementations.

For example, if you want to use all standard implementations of all primitives in the current release of Tink, the initialization would look as follows:

import tink
from tink import tink_config
tink_config.register()

To use standard implementations of only one primitive, say AEAD:

import tink
from tink import aead
aead.register()

The registration of custom key managers can proceed directly via the core.Registry class:

import tink
from tink import core
core.Registry.register_key_manager(CustomAeadKeyManager())

Generating new keys and keysets

Each KeyManager implementation provides a new_key_data(key_template) method that generates new keys of the corresponding key type. However, to avoid accidental leakage of sensitive key material you should avoid mixing key(set) generation with key(set) usage in code.

To support the separation between these activities Tink package provides a command-line tool called Tinkey, which can be used for common key management tasks.

Still, if there is a need to generate a KeysetHandle with fresh key material directly in Python code, you can use tink.new_keyset_handle:

import tink
from tink import aead

key_template = aead.aead_key_templates.AES128_EAX
keyset_handle = tink.new_keyset_handle(key_template)
# use the keyset...

where key_template can be obtained from util classes corresponding to Tink primitives, e.g. mac_key_templates, aead_key_templates, or hybrid_key_templates.

Loading existing keysets

To load encrypted keysets, use tink.read_keyset_handle and an appropriate KeysetReader, depending on the wire format of the stored keyset, for example a tink.BinaryKeysetReader or a tink.JsonKeysetReader.

import tink

json_encrypted_keyset = ...
reader = tink.JsonKeysetReader(json_encrypted_keyset)
keyset_handle = tink.read_keyset_handle(reader, master_key_aead)

To load cleartext keysets, use cleartext_keyset_handle and an appropriate KeysetReader.

import tink
from tink import cleartext_keyset_handle

json_keyset = ...
reader = tink.JsonKeysetReader(json_keyset)
keyset_handle = cleartext_keyset_handle.read(reader)

Obtaining and using primitives

Primitives represent cryptographic operations offered by Tink, hence they form the core of the Tink API. A primitive is just an interface that specifies what operations are offered by the primitive. A primitive can have multiple implementations, and you choose a desired implementation by using a key of corresponding type (see this section for further details).

Tink for Python supports the same primitives as Tink for C++. A list of primitives and their implementations currently supported by Tink in C++ can be found here.

You obtain a primitive by calling the method primitive of the KeysetHandle.

Symmetric key encryption

You can obtain and use an AEAD (Authenticated Encryption with Associated Data) primitive to encrypt or decrypt data:

import tink
from tink import aead

plaintext = b'your data...'
associated_data = b'context'

# Register all AEAD primitives
aead.register()

# 1. Get a handle to the key material.
keyset_handle = tink.new_keyset_handle(aead.aead_key_templates.AES256_GCM)

# 2. Get the primitive.
aead_primitive = keyset_handle.primitive(aead.Aead)

# 3. Use the primitive.
ciphertext = aead_primitive.encrypt(plaintext, associated_data)

Deterministic symmetric key encryption

You can obtain and use a DeterministicAEAD (Deterministic Authenticated Encryption with Associated Data primitive to encrypt or decrypt data:

import tink
from tink import daead

plaintext = b'your data...'
associated_data = b'context'

# Register all deterministic AEAD primitives
daead.register()

# 1. Get a handle to the key material.
keyset_handle = tink.new_keyset_handle(daead.deterministic_aead_key_templates.AES256_SIV)

# 2. Get the primitive.
daead_primitive = keyset_handle.primitive(daead.DeterministicAead)

# 3. Use the primitive.
ciphertext = daead_primitive.encrypt_deterministically(plaintext, associated_data)

Symmetric key encryption of streaming data

You can obtain and use a Streaming AEAD (Streaming Authenticated Encryption with Associated Data) primitive to encrypt or decrypt data streams:

import tink
from tink import streaming_aead

long_plaintext = b'your data...'
associated_data = b'context'

# Register all streaming AEAD primitives
streaming_aead.register()

# 1. Get a handle to the key material
keyset_handle = tink.new_keyset_handle(streaming_aead.streaming_aead_key_templates.AES256_CTR_HMAC_SHA256_4KB)

# 2. Get the primitive.
streaming_aead_primitive = keyset_handle.primitive(streaming_aead.StreamingAead)

# 3. Use the primitive.
output_file = open("ciphertext.out", 'wb')
with streaming_aead_primitive.new_encrypting_stream(output_file, associated_data) as enc_stream:
  bytes_written = enc_stream.write(long_plaintext)

Message Authentication Code

You can compute or verify a MAC (Message Authentication Code):

import tink
from tink import mac

data = b'your data...'

# Register all MAC primitives
mac.register()

# 1. Get a handle to the key material.
keyset_handle = tink.new_keyset_handle(mac.mac_key_templates.HMAC_SHA256_128BITTAG)

# 2. Get the primitive.
mac_primitive = keyset_handle.primitive(mac.Mac)

# 3. Use the primitive to compute a tag,
tag = mac_primitive.compute_mac(data)

# ... or to verify a tag.
mac_primitive.verify_mac(tag, data)

Hybrid Encryption

To encrypt or decrypt using a combination of public key encryption and symmetric key encryption one can use the following:

import tink
from tink import hybrid

plaintext = b'your data...'
context = b'context'

# Register all Hybrid primitives
hybrid.register()

# 1. Generate the private key material.
private_keyset_handle = tink.new_keyset_handle(hybrid.hybrid_key_templates.ECIES_P256_HKDF_HMAC_SHA256_AES128_GCM)

# Obtain the public key material.
public_keyset_handle = private_keyset_handle.public_keyset_handle()

# Encryption

# 2. Get the primitive.
hybrid_encrypt = public_keyset_handle.primitive(hybrid.HybridEncrypt)

# 3. Use the primitive.
ciphertext = hybrid_encrypt.encrypt(plaintext, context)

# Decryption

# 2. Get the primitive.
hybrid_decrypt = private_keyset_handle.primitive(hybrid.HybridDecrypt)

# 3. Use the primitive.
plaintext = hybrid_decrypt.decrypt(ciphertext, context)

Digital Signatures

You can sign or verify using a digital signature:

import tink
from tink import signature

# Register key manager for signatures
signature.register()

# Signing
# 1. Generate the private key material.
keyset_handle = tink.new_keyset_handle(signature.signature_key_templates.ED25519)

# 2. Get the primitive.
signer = keyset_handle.primitive(signature.PublicKeySign)

# 3. Use the primitive to sign.
signature_data = signer.sign(b'your data')

# Verifying
# 1. Obtain the public key material.
public_keyset_handle = keyset_handle.public_keyset_handle()

# 2. Get the primitive.
verifier = public_keyset_handle.primitive(signature.PublicKeyVerify)

# 3. Use the primitive to verify.
verifier.verify(signature_data, b'your data')

Envelope encryption

Via the AEAD interface, Tink supports envelope encryption.

For example, you can perform envelope encryption with a Google Cloud KMS key at gcp-kms://projects/tink-examples/locations/global/keyRings/foo/cryptoKeys/bar using the credentials in credentials.json as follows:

import tink
from tink import aead
from tink.integration import gcpkms

key_uri = 'gcp-kms://projects/tink-examples/locations/global/keyRings/foo/cryptoKeys/bar'
gcp_credentials = 'credentials.json'

plaintext = b'your data...'
associated_data = b'context'

# Read the GCP credentials and setup client
try:
  gcp_client = gcpkms.GcpKmsClient(key_uri, gcp_credentials)
  gcp_aead = gcp_client.get_aead(key_uri)
except tink.TinkError as e:
  logging.error('Error initializing GCP client: %s', e)
  return 1

# Create envelope AEAD primitive using AES256 GCM for encrypting the data
try:
  key_template = aead.aead_key_templates.AES256_GCM
  env_aead = aead.KmsEnvelopeAead(key_template, gcp_aead)
except tink.TinkError as e:
  logging.error('Error creating primitive: %s', e)
  return 1

# Use env_aead to encrypt data
ciphertext = env_aead.encrypt(plaintext, associated_data)