Secure machine learning analytics using homomorphic encryption
Provided are methods and systems for performing a secure machine learning analysis over an instance of data. An example method includes acquiring, by a client, an homomorphic encryption scheme, and at least one machine learning model data structure. The method further includes generating, using the encryption scheme, at least one homomorphically encrypted data structure, and sending the encrypted data structure to at least one server. The method includes executing a machine learning model, by the at least one server based on the encrypted data structure to obtain an encrypted result. The method further includes sending, by the server, the encrypted result to the client where the encrypted result is decrypted. The machine learning model includes neural networks and decision trees.
1. A method for performing a secure machine learning analysis using homomorphic encryption, the method comprising:
receiving, from a client, by at least one server, the encrypted at least one machine learning data structure, the at least one machine learning data structure being encrypted using a fully homomorphic encryption scheme;
extracting, by the at least one server, an instance;
evaluating, by the at least one server, the at least one machine learning data structure over the instance utilizing a trained machine learning model to obtain at least one encrypted result; and
sending, by the at least one server, the at least one encrypted result to the client, wherein the client is configured to decrypt the at least one encrypted result using the homomorphic encryption scheme,
wherein the at least one machine learning data structure is generated based on a trained machine learning model, the trained machine learning model is a decision tree, the at least one machine learning data structure includes at least one feature vector, the decision tree has a decision depth, and the at least one feature vector has a feature depth, and wherein the feature depth is greater than the decision depth.
2. The method of claim 1 , wherein the homomorphic encryption scheme includes a fully homomorphic encryption scheme.
3. The method of claim 1 , wherein the homomorphic encryption scheme includes at least one of a Brakerski/Fan-Vercauteren and a Cheon-Kim-Kim-Song cryptosystem.
4. The method of claim 1 , wherein the evaluating, by the at least one server, is performed in an unsecure environment.
5. The method of claim 1 , wherein the trained machine learning model is trained in a trusted environment.
6. The method of claim 1 , wherein the at least one feature vector consist of binary values.
7. A system for performing a secure machine learning analysis using homomorphic encryption, the system comprising:
at least one processor; and
a memory communicatively coupled with the at least one processor, the memory storing instructions, which when executed by the at least one processor perform a method comprising:
receiving, from a client, at least one machine learning data, the at least machine learning data structure being encrypted using a homomorphic encryption scheme;
extracting, by at least one server, an instance of data;
evaluating, by the at least one server, the at least one machine learning data structure over the instance utilizing a trained machine learning model to obtain at least one encrypted results; and
sending the at least one encrypted results to the client, wherein the client is configured to decrypt the at least one encrypted results using the homomorphic encryption scheme,
wherein the client is configured with at least one machine learning data structure based on training a machine learning model, wherein the machine learning model is a decision tree wherein the at least one machine learning data structure includes at least one feature vector, the decision tree has a decision depth, the at least one feature vector has a feature depth, and wherein the feature depth is greater than the decision depth.
8. The system of claim 7 , wherein the homomorphic encryption scheme includes a fully homomorphic encryption scheme.
9. The system of claim 7 , wherein the homomorphic encryption scheme includes at least one of a Brakerski/Fan-Vercauteren and a Cheon-Kim-Kim-Song cryptosystem.
10. The system of claim 7 , wherein the at least one feature vector consists of binary values.
11. The system of claim 7 , wherein the evaluating, by the at least one server, is performed in an unsecure environment.
12. The system of claim 7 , wherein the trained machine learning model is trained in a trusted environment.
13. A non-transitory computer-readable storage medium having embodied thereon instructions, which when executed by at least one processor, perform steps of a method, the method comprising:
receiving, from a client, by at least one server, at least one encrypted machine learning data structure, the at least one encrypted machine learning data structure being encrypted using a fully homomorphic encryption scheme;
extracting, by the at least one server, an instance;
evaluating, by the at least one server, the at least one encrypted machine learning data structure over the instance utilizing a trained machine learning model to obtain at least one encrypted result; and
sending, by the at least one server, the at least one encrypted result to the client, wherein the client is configured to decrypt the at least one encrypted result using the homomorphic encryption scheme,
wherein the at least one encrypted machine learning data structure is generated based on a trained machine learning model, the trained machine learning model is a decision tree, the at least one encrypted machine learning data structure includes at least one feature vector, the decision tree has a decision depth, and the at least one feature vector has a feature depth, and wherein the feature depth is greater than the decision depth.
14. The non-transitory computer-readable storage medium of claim 13 , wherein the homomorphic encryption scheme includes a fully homomorphic encryption scheme.
15. The non-transitory computer-readable storage medium of claim 13 , wherein the homomorphic encryption scheme includes at least one of a Brakerski/Fan-Vercauteren and a Cheon-Kim-Kim-Song cryptosystem.
16. The non-transitory computer-readable storage medium of claim 13 , wherein the trained machine learning model is trained in a trusted environment.
17. The non-transitory computer-readable storage medium of claim 13 , wherein the evaluating, by the at least one server, is performed in an unsecure environment.
18. The non-transitory computer-readable storage medium of claim 13 , wherein the at least one feature vector consists of binary values.