The AWS Encryption SDK for Python provides a fully compliant, native Python implementation of the AWS Encryption SDK.

The latest full documentation can be found at Read the Docs.

Find us on GitHub.

Security issue notifications

See Support Policy for details on the current support status of all major versions of this library.

• Python 3.7+
• cryptography >= 3.4.6
• boto3 >= 1.10.0
• attrs

$ pip install aws-encryption-sdk

There are four main concepts that you need to understand to use this library:

Cryptographic materials managers (CMMs) are resources that collect cryptographic materials and prepare them for use by the Encryption SDK core logic.

An example of a CMM is the default CMM, which is automatically generated anywhere a caller provides a master key provider. The default CMM collects encrypted data keys from all master keys referenced by the master key provider.

An example of a more advanced CMM is the caching CMM, which caches cryptographic materials provided by another CMM.

Master key providers are resources that provide master keys. An example of a master key provider is AWS KMS.

To encrypt data in this client, a MasterKeyProvider object must contain at least one MasterKey object.

MasterKeyProvider objects can also contain other MasterKeyProvider objects.

Master keys generate, encrypt, and decrypt data keys. An example of a master key is a KMS customer master key (CMK).

Data keys are the encryption keys that are used to encrypt your data. If your algorithm suite uses a key derivation function, the data key is used to generate the key that directly encrypts the data.

To use this module, you (the caller) must first create an instance of the EncryptionSDKClient class. The constructor to this class accepts an optional keyword argument, commitment_policy, that controls which algorithm suites can be used for encryption and decryption. If no value is provided for this argument, a default value of REQUIRE_ENCRYPT_REQUIRE_DECRYPT is used. Unless you have specialized performance requirements or are in the process of migrating from an older version of the AWS Encryption SDK, we recommend using the default value.

import aws_encryption_sdk
from aws_encryption_sdk.identifiers import CommitmentPolicy


client = aws_encryption_sdk.EncryptionSDKClient(
    commitment_policy=CommitmentPolicy.REQUIRE_ENCRYPT_REQUIRE_DECRYPT
)

You must then create an instance of either a master key provider or a CMM. The examples in this readme use the StrictAwsKmsMasterKeyProvider class.

A StrictAwsKmsMasterKeyProvider is configured with an explicit list of AWS KMS CMKs with which to encrypt and decrypt data. On encryption, it encrypts the plaintext with all configured CMKs. On decryption, it only attempts to decrypt ciphertexts that have been wrapped with a CMK that matches one of the configured CMK ARNs.

To create a StrictAwsKmsMasterKeyProvider you must provide one or more CMKs. For providers that will only be used for encryption, you can use any valid KMS key identifier. For providers that will be used for decryption, you must use the key ARN; key ids, alias names, and alias ARNs are not supported.

Because the StrictAwsKmsMasterKeyProvider uses the boto3 SDK to interact with AWS KMS, it requires AWS Credentials. To provide these credentials, use the standard means by which boto3 locates credentials or provide a pre-existing instance of a botocore session to the StrictAwsKmsMasterKeyProvider. This latter option can be useful if you have an alternate way to store your AWS credentials or you want to reuse an existing instance of a botocore session in order to decrease startup costs.

If you configure the the StrictAwsKmsMasterKeyProvider with multiple CMKs, the final message will include a copy of the data key encrypted by each configured CMK.

import aws_encryption_sdk

kms_key_provider = aws_encryption_sdk.StrictAwsKmsMasterKeyProvider(key_ids=[
    'arn:aws:kms:us-east-1:2222222222222:key/22222222-2222-2222-2222-222222222222',
    'arn:aws:kms:us-east-1:3333333333333:key/33333333-3333-3333-3333-333333333333'
])

You can add CMKs from multiple regions to the StrictAwsKmsMasterKeyProvider.

import aws_encryption_sdk

kms_key_provider = aws_encryption_sdk.StrictAwsKmsMasterKeyProvider(key_ids=[
    'arn:aws:kms:us-east-1:2222222222222:key/22222222-2222-2222-2222-222222222222',
    'arn:aws:kms:us-west-2:3333333333333:key/33333333-3333-3333-3333-333333333333',
    'arn:aws:kms:ap-northeast-1:4444444444444:key/44444444-4444-4444-4444-444444444444'
])

We recommend using a StrictAwsKmsMasterKeyProvider in order to ensure that you can only encrypt and decrypt data using the AWS KMS CMKs you expect. However, if you are unable to explicitly identify the AWS KMS CMKs that should be used for decryption, you can instead use a DiscoveryAwsKmsMasterKeyProvider for decryption operations. This provider attempts decryption of any ciphertexts as long as they match a DiscoveryFilter that you configure. A DiscoveryFilter consists of a list of AWS account ids and an AWS partition.

import aws_encryption_sdk
from aws_encryption_sdk.key_providers.kms import DiscoveryFilter

discovery_filter = DiscoveryFilter(
    account_ids=['222222222222', '333333333333'],
    partition='aws'
)
kms_key_provider = aws_encryption_sdk.DiscoveryAwsKmsMasterKeyProvider(
    discovery_filter=discovery_filter
)

If you do not want to filter the set of allowed accounts, you can also omit the discovery_filter argument.

Note that a DiscoveryAwsKmsMasterKeyProvider cannot be used for encryption operations.

After you create an instance of an EncryptionSDKClient and a MasterKeyProvider, you can use either of the client's two encrypt/decrypt functions to encrypt and decrypt your data.

import aws_encryption_sdk
from aws_encryption_sdk.identifiers import CommitmentPolicy

client = aws_encryption_sdk.EncryptionSDKClient(
    commitment_policy=CommitmentPolicy.FORBID_ENCRYPT_ALLOW_DECRYPT
)

kms_key_provider = aws_encryption_sdk.StrictAwsKmsMasterKeyProvider(key_ids=[
    'arn:aws:kms:us-east-1:2222222222222:key/22222222-2222-2222-2222-222222222222',
    'arn:aws:kms:us-east-1:3333333333333:key/33333333-3333-3333-3333-333333333333'
])
my_plaintext = b'This is some super secret data!  Yup, sure is!'

my_ciphertext, encryptor_header = client.encrypt(
    source=my_plaintext,
    key_provider=kms_key_provider
)

decrypted_plaintext, decryptor_header = client.decrypt(
    source=my_ciphertext,
    key_provider=kms_key_provider
)

assert my_plaintext == decrypted_plaintext
assert encryptor_header.encryption_context == decryptor_header.encryption_context

You can provide an encryption context: a form of additional authenticating information.

import aws_encryption_sdk
from aws_encryption_sdk.identifiers import CommitmentPolicy

client = aws_encryption_sdk.EncryptionSDKClient(
    commitment_policy=CommitmentPolicy.FORBID_ENCRYPT_ALLOW_DECRYPT
)

kms_key_provider = aws_encryption_sdk.StrictAwsKmsMasterKeyProvider(key_ids=[
    'arn:aws:kms:us-east-1:2222222222222:key/22222222-2222-2222-2222-222222222222',
    'arn:aws:kms:us-east-1:3333333333333:key/33333333-3333-3333-3333-333333333333'
])
my_plaintext = b'This is some super secret data!  Yup, sure is!'

my_ciphertext, encryptor_header = client.encrypt(
    source=my_plaintext,
    key_provider=kms_key_provider,
    encryption_context={
        'not really': 'a secret',
        'but adds': 'some authentication'
    }
)

decrypted_plaintext, decryptor_header = client.decrypt(
    source=my_ciphertext,
    key_provider=kms_key_provider
)

assert my_plaintext == decrypted_plaintext
assert encryptor_header.encryption_context == decryptor_header.encryption_context

If you are handling large files or simply do not want to put the entire plaintext or ciphertext in memory at once, you can use this library's streaming clients directly. The streaming clients are file-like objects, and behave exactly as you would expect a Python file object to behave, offering context manager and iteration support.

import aws_encryption_sdk
from aws_encryption_sdk.identifiers import CommitmentPolicy
import filecmp

client = aws_encryption_sdk.EncryptionSDKClient(
    commitment_policy=CommitmentPolicy.FORBID_ENCRYPT_ALLOW_DECRYPT
)

kms_key_provider = aws_encryption_sdk.StrictAwsKmsMasterKeyProvider(key_ids=[
    'arn:aws:kms:us-east-1:2222222222222:key/22222222-2222-2222-2222-222222222222',
    'arn:aws:kms:us-east-1:3333333333333:key/33333333-3333-3333-3333-333333333333'
])
plaintext_filename = 'my-secret-data.dat'
ciphertext_filename = 'my-encrypted-data.ct'

with open(plaintext_filename, 'rb') as pt_file, open(ciphertext_filename, 'wb') as ct_file:
    with client.stream(
        mode='e',
        source=pt_file,
        key_provider=kms_key_provider
    ) as encryptor:
        for chunk in encryptor:
            ct_file.write(chunk)

new_plaintext_filename = 'my-decrypted-data.dat'

with open(ciphertext_filename, 'rb') as ct_file, open(new_plaintext_filename, 'wb') as pt_file:
    with client.stream(
        mode='d',
        source=ct_file,
        key_provider=kms_key_provider
    ) as decryptor:
        for chunk in decryptor:
            pt_file.write(chunk)

assert filecmp.cmp(plaintext_filename, new_plaintext_filename)
assert encryptor.header.encryption_context == decryptor.header.encryption_context

Adjusting the frame size can significantly improve the performance of encrypt/decrypt operations with this library.

Processing each frame in a framed message involves a certain amount of overhead. If you are encrypting a large file, increasing the frame size can offer potentially significant performance gains. We recommend that you tune these values to your use-case in order to obtain peak performance.

The EncryptionSDKClient and all provided CryptoMaterialsManager are thread safe. But instances of BaseKMSMasterKeyProvider MUST not be shared between threads, for the reasons outlined in the boto3 docs.

Because the BaseKMSMaterKeyProvider creates a new boto3 sessions per region, users do not need to create a client for every region in every thread; a new BaseKMSMasterKeyProvider per thread is sufficient.

(The BaseKMSMasterKeyProvider is the internal parent class of all the KMS Providers.)

Finally, while the CryptoMaterialsCache is thread safe, sharing entries in that cache across threads needs to be done carefully (see the !Note about partition name in the API Docs).