Using the DynamoDB Encryption Client for Java
Note
Our client-side encryption library was renamed to AWS Database Encryption SDK. The following topic provides information on versions 1.x—2.x of the DynamoDB Encryption Client for Java and versions 1.x—3.x of the DynamoDB Encryption Client for Python. For more information, see AWS Database Encryption SDK for DynamoDB version support.
This topic explains some of the features of the DynamoDB Encryption Client in Java that might not be found in other programming language implementations.
For details about programming with the DynamoDB Encryption Client, see the Java
examples, the examplesaws-dynamodb-encryption-java repository
on GitHub, and the Javadoc
Item encryptors: AttributeEncryptor and DynamoDBEncryptor
The DynamoDB Encryption Client in Java has two item encryptors: the
lower-level DynamoDBEncryptor
The AttributeEncryptor
is a helper class that helps you use the DynamoDBMapper in the AWS SDK for Java with the DynamoDB Encryptor
in the DynamoDB Encryption Client. When you use the AttributeEncryptor
with the DynamoDBMapper
, it transparently encrypts and signs your items when you save them. It also transparently verifies and decrypts
your items when you load them.
Configuring save behavior
You can use the AttributeEncryptor
and DynamoDBMapper
to add or
replace table items with attributes that are signed only or encrypted and signed. For these
tasks, we recommend that you configure it to use the PUT
save behavior, as shown
in the following example. Otherwise, you might not be able to decrypt your data.
DynamoDBMapperConfig mapperConfig = DynamoDBMapperConfig.builder().withSaveBehavior(SaveBehavior.PUT).build(); DynamoDBMapper mapper = new DynamoDBMapper(ddb, mapperConfig, new AttributeEncryptor(encryptor));
If you use the default save behavior, which updates only the attributes that are modeled in the table item, attributes that are not modeled are not included in the signature, and are not changed by table writes. As a result, on later reads of all attributes, the signature will not validate, because it doesn't include un-modeled attributes.
You can also use the CLOBBER
save behavior. This behavior is identical to the
PUT
save behavior except that it disables optimistic locking and overwrites the
item in the table.
To prevent signature errors, the DynamoDB Encryption Client throws a runtime exception if an
AttributeEncryptor
is used with a DynamoDBMapper
that is not
configured with a save behavior of CLOBBER
or PUT
.
To see this code used in an example, see Using the DynamoDBMapper and the AwsKmsEncryptedObject.javaaws-dynamodb-encryption-java
repository in GitHub.
Attribute actions in Java
Attribute actions determine which attribute
values are encrypted and signed, which are only signed, and which are ignored. The method you
use to specify attribute actions depends on whether you use the DynamoDBMapper
and AttributeEncryptor
, or the lower-level DynamoDBEncryptor
Important
After you use your attribute actions to encrypt your table items, adding or removing attributes from your data model might cause a signature validation error that prevents you from decrypting your data. For a detailed explanation, see Changing your data model.
When you use the DynamoDBMapper
and AttributeEncryptor
, you use annotations to specify the attribute
actions. The DynamoDB Encryption Client uses the standard DynamoDB attribute annotations that define
the attribute type to determine how to protect an attribute. By default, all attributes are encrypted and signed except for primary keys, which
are signed but not encrypted.
Note
Do not encrypt the value of attributes with the @DynamoDBVersionAttribute annotation, although you can (and should) sign them. Otherwise, conditions that use its value will have unintended effects.
// Attributes are encrypted and signed @DynamoDBAttribute(attributeName="Description") // Partition keys are signed but not encrypted @DynamoDBHashKey(attributeName="Title") // Sort keys are signed but not encrypted @DynamoDBRangeKey(attributeName="Author")
To specify exceptions, use the encryption annotations defined in the DynamoDB Encryption Client for Java. If you specify them at the class level, they become the default value for the class.
// Sign only @DoNotEncrypt // Do nothing; not encrypted or signed @DoNotTouch
For example, these annotations sign but do not encrypt the
PublicationYear
attribute, and do not encrypt or sign the ISBN
attribute value.
// Sign only (override the default) @DoNotEncrypt @DynamoDBAttribute(attributeName="PublicationYear") // Do nothing (override the default) @DoNotTouch @DynamoDBAttribute(attributeName="ISBN")
To specify attribute actions when you use the DynamoDBEncryptorHashMap
object in which the
name-value pairs represent attribute names and the specified actions.
The valid values are for the attribute actions are defined in the
EncryptionFlags
enumerated type. You can use ENCRYPT
and
SIGN
together, use SIGN
alone, or omit both. However, if you use
ENCRYPT
alone, the DynamoDB Encryption Client throws an error. You cannot encrypt an attribute
that you don't sign.
ENCRYPT SIGN
Warning
Do not encrypt the primary key attributes. They must remain in plaintext so DynamoDB can find the item without running a full table scan.
If you specify a primary key in the encryption context and then specify ENCRYPT
in the attribute action for either primary key attribute, the DynamoDB Encryption Client throws an
exception.
For example, the following Java code creates an actions
HashMap that
encrypts and signs all attributes in the record
item. The exceptions are the
partition key and sort key attributes, which are signed but not encrypted, and the
test
attribute, which is not signed or encrypted.
final EnumSet<EncryptionFlags> signOnly = EnumSet.of(EncryptionFlags.SIGN); final EnumSet<EncryptionFlags> encryptAndSign = EnumSet.of(EncryptionFlags.ENCRYPT, EncryptionFlags.SIGN); final Map<String, Set<EncryptionFlags>> actions = new HashMap<>(); for (final String attributeName : record.keySet()) { switch (attributeName) { case partitionKeyName: // no break; falls through to next case case sortKeyName: // Partition and sort keys must not be encrypted, but should be signed actions.put(attributeName, signOnly); break; case "test": // Don't encrypt or sign break; default: // Encrypt and sign everything else actions.put(attributeName, encryptAndSign); break; } }
Then, when you call the encryptRecordDynamoDBEncryptor
, specify the map as
the value of the attributeFlags
parameter. For example, this call to
encryptRecord
uses the actions
map.
// Encrypt the plaintext record final Map<String, AttributeValue> encrypted_record = encryptor.encryptRecord(record, actions, encryptionContext);
Overriding table names
In the DynamoDB Encryption Client, the name of the DynamoDB table is an element of the DynamoDB encryption context that is passed to the encryption and decryption methods. When you encrypt or sign table items, the DynamoDB encryption context, including the table name, is cryptographically bound to the ciphertext. If the DynamoDB encryption context that is passed to the decrypt method doesn't match the DynamoDB encryption context that was passed to the encrypt method, the decrypt operation fails.
Occasionally, the name of a table changes, such as when you back up a table or perform a point-in-time recovery. When you decrypt or verify the signature of these items, you must pass in the same DynamoDB encryption context that was used to encrypt and sign the items, including the original table name. The current table name is not needed.
When you use the DynamoDBEncryptor
, you assemble the DynamoDB encryption context
manually. However, if you are using the DynamoDBMapper
, the
AttributeEncryptor
creates the DynamoDB encryption context for you, including the
current table name. To tell the AttributeEncryptor
to create an encryption
context with a different table name, use the EncryptionContextOverrideOperator
.
For example, the following code creates instances of the cryptographic materials provider (CMP) and the DynamoDBEncryptor
. Then it calls
the setEncryptionContextOverrideOperator
method of the DynamoDBEncryptor
. It uses the
overrideEncryptionContextTableName
operator, which overrides one table name. When it is configured this way, the
AttributeEncryptor
creates a DynamoDB encryption context that includes newTableName
in place of oldTableName
.
For a complete example, see EncryptionContextOverridesWithDynamoDBMapper.java
final DirectKmsMaterialProvider cmp = new DirectKmsMaterialProvider(kms, keyArn); final DynamoDBEncryptor encryptor = DynamoDBEncryptor.getInstance(cmp); encryptor.setEncryptionContextOverrideOperator(EncryptionContextOperators.overrideEncryptionContextTableName( oldTableName, newTableName));
When you call the load method of the DynamoDBMapper
, which decrypts and verifies the item, you specify the original table name.
mapper.load(itemClass, DynamoDBMapperConfig.builder() .withTableNameOverride(DynamoDBMapperConfig.TableNameOverride.withTableNameReplacement(oldTableName)) .build());
You can also use the overrideEncryptionContextTableNameUsingMap
operator,
which overrides multiple table names.
The table name override operators are typically used when decrypting data and verifying signatures. However, you can use them to set the table name in the DynamoDB encryption context to a different value when encrypting and signing.
Do not use the table name override operators if you are using the DynamoDBEncryptor
. Instead, create an encryption context with the
original table name and submit it to the decryption method.