Rosetta v1.0.0

1. Support multi-tasking concurrency, that mean different privacy protocols can be executed simultaneously.

2. Nodes can be configured with roles, supporting calculation node, data node and result node, and data nodes can be configured in any number according to requirements.

3. Refactoring code to make the code highly cohesive and low coupled.

4. Added SecureExp, SecureSqrt, SecureRsqrt, SecureReciprocaldiv secure operations.

5. Adding audit logs, which are turned off by default and can be turned on according to business scenarios.

6. Efficient third-party io modules can be integrated into Rosetta by implementing specific interfaces.

7. Optimize back-end unit tests to provide more friendly test cases.

8. Some known bugs are fixed.

Rosetta v0.3.0

1. Added PrivateTextLineDataset, PrivateInput secure operations.
2. Added SecureLogicalAnd, SecureLogicalOr, SecureLogicalXor, SecureLogicalNot secure operations.
3. Speedup some backend operations.
4. Uses related python classes such as PrivateTextLineDataset and iterators to load large data sets, thereby reducing memory usage.
5. Some known bugs are fixed.

Rosetta v0.2.0

1. Refactor the whole architecture so that each layer has a clearer boundary. Especially, an abstract protocol layer is added to decouple cryptographic protocols and TensorFlow Ops, so that TensorFlow does not depend on specific protocols and vise verse.

2. Support custom data types by mainly utilizing tf.string. This will enable us to have arbitrary length and format of internal ciphertext.

3. Some APIs, such as activate, are added and refined to support better control on selection and configuration of backend protocol. In this version, you can configure IP ports, host addresses, save mode flag and protocol precision as JSON string by calling activate in your program.

4. Some known bugs are fixed.

Rosetta v0.1.1

Rosetta v0.1.1

• Binary installation of TensorFlow is supported

Rosetta v0.1.0

Rosetta v0.1.0