Analytic workload partitioning for security and performance optimization

Patent metadata

Field	Value
Publication number	US-11567761-B2
Application number	US-202016912148-A
Country	US
Kind code	B2
Filing date	Jun 25, 2020
Priority date	Jun 25, 2020
Publication date	Jan 31, 2023
Grant date	Jan 31, 2023

Abstract

The present disclosure provides privacy preservation of analytic workflows based on splitting the workflow into sub-workflows each with different privacy-preserving characteristics. Libraries are generated that provide for formatting and/or encrypting data for use in the sub-workflows and also for compiling a machine learning algorithm for the sub-workflows. Subsequently, the sub-workflows can be executed using the compiled algorithm and formatted data.

First claim

What is claimed is:

1. A computer-implemented method, comprising:
receiving, at a processing unit, a first information element comprising indications of information about data for an analytic workflow, the data to comprise homomorphically encrypted data which allows application of mathematical operands directly to the homomorphically encrypted data without decryption of the homomorphically encrypted data;
receiving, at the processing unit, a second information element comprising indication of information about the analytic workflow;
splitting the analytic workflow into a plurality of sub-workflows based on the information about the data and the information about the analytic workflow, wherein one of the plurality of sub-workflows corresponds to a homomorphic encrypted data operation;
generating a data format library comprising indications of at least one data format for the sub-workflows; and
generating an algorithm compilation library comprising indications of the sub-workflows.

2. The computer-implemented method of claim 1 , wherein the information about the analytic workflow comprises indications of mathematical primitives used in the analytic workflow.

3. The computer-implemented method of claim 1 , wherein the data format library comprises indications of a data format and a data encryption scheme.

4. The computer-implemented method of claim 1 , wherein the algorithm compilation library comprises indications of a first sub-workflow comprising and a second sub-workflow different from the first sub-workflow.

5. The computer-implemented method of claim 4 , wherein the first sub-workflow comprises execution of a first mathematical primitive on homomorphic encrypted data and the second sub-workflow comprises execution of a second mathematical primitive on decrypted data, wherein the first mathematical primitive is different from the second mathematical primitive.

6. The computer-implemented method of claim 1 , comprising:
receiving, at the processing unit, a third information element comprising indication of information about communication requirements of the analytic workflow; and
splitting the analytic workflow into the plurality of sub-workflows further based on the information about the communication requirements of the analytic workflow.

7. The computer-implemented method of claim 1 , comprising:
sending the data format library to a data owner device;
receiving, from the data owner device, formatted and encrypted data based on the data format library;
sending the algorithm compilation library to an analytic workflow owner device;
receiving, from the analytic workflow owner device, a compiled machine learning model based on the algorithm compilation library, wherein the compiled machine learning model comprises the plurality of sub-workflows; and
executing the compiled machine learning model with the formatted and encrypted data.

8. A computing apparatus, the computing apparatus comprising:
a processor; and
memory storing instructions, which when executed by the processor, cause the apparatus to:
receive a first information element comprising indications of information about data for an analytic workflow, the data to comprise homomorphically encrypted data which allows application of mathematical operands directly to the homomorphically encrypted data without decryption of the homomorphically encrypted data;
receive a second information element comprising indication of information about the analytic workflow;
split the analytic workflow into a plurality of sub-workflows based on the information about the data and the information about the analytic workflow, wherein one of the plurality of sub-workflows corresponds to a homomorphic encrypted data operation;
generate a data format library comprising indications of at least one data format for the sub-workflows; and
generate an algorithm compilation library comprising indications of the sub-workflows.

9. The computing apparatus of claim 8 , wherein the information about the analytic workflow comprises indications of mathematical primitives used in the analytic workflow.

10. The computing apparatus of claim 8 , wherein the data format library comprises indications of a data format and a data encryption scheme.

11. The computing apparatus of claim 8 , wherein the algorithm compilation library comprises indications of a first sub-workflow comprising and a second sub-workflow different from the first sub-workflow.

12. The computing apparatus of claim 11 , wherein the first sub-workflow comprises execution of a first mathematical primitive on homomorphic encrypted data and the second sub-workflow comprises execution of a second mathematical primitive on decrypted data, wherein the first mathematical primitive is different from the second mathematical primitive.

13. The computing apparatus of claim 8 , the instructions when executed by the processor, cause the apparatus to:
receive, a third information element comprising indication of information about communication requirements of the analytic workflow; and
split the analytic workflow into the plurality of sub-workflows further based on the information about the communication requirements of the analytic workflow.

14. The computing apparatus of claim 8 , the instructions when executed by the processor, cause the apparatus to:
send the data format library to a data owner device; and
receive, from the data owner device, formatted and encrypted data based on the data format library.

15. The computing apparatus of claim 14 , the instructions when executed by the processor, cause the apparatus to:
send the algorithm compilation library to an analytic workflow owner device; and
receive, from the analytic workflow owner device, a compiled machine learning model based on the algorithm compilation library, wherein the compiled machine learning model comprises the plurality of sub-workflows.

16. The computing apparatus of claim 15 , the instructions when executed by the processor, cause the apparatus to execute the compiled machine learning model with the formatted and encrypted data.

17. A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a computer, cause the computer to:
receive, at a processing unit, a first information element comprising indications of information about data for an analytic workflow, the data to comprise homomorphically encrypted data which allows application of mathematical operands directly to the homomorphically encrypted data without decryption of the homomorphically encrypted data;
receive, at the processing unit, a second information element comprising indication of information about the analytic workflow;
split the analytic workflow into a plurality of sub-workflows based on the information about the data and the information about the analytic workflow, wherein one of the plurality of sub-workflows corresponds to a homomorphic encrypted data operation;
generate a data format library comprising indications of at least one data format for the sub-workflows; and
generate an algorithm compilation library comprising indications of the sub-workflows.

18. The computer-readable storage medium of claim 17 , wherein the information about the analytic workflow comprises indications of mathematical primitives used in the analytic workflow.

19. The computer-readable storage medium of claim 17 , wherein the data format library comprises indications of a data format and a data encryption scheme.

20.

Assignees

• Intel Corp

Inventors

• Levin Itamar
• Melzer Guilad
• Nayshtut Alex
• Kellerman Raizy

Classifications

• G06F21/52

  during program execution, e.g. stack integrity {; Preventing unwanted data erasure; Buffer overflow} · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F21/55

  Detecting local intrusion or implementing counter-measures · CPC title

• G06F21/6209

  to a single file or object, e.g. in a secure envelope, encrypted and accessed using a key, or with access control rules appended to the object itself · CPC title

• G06F9/226 Primary

  Microinstruction function, e.g. input/output microinstruction; diagnostic microinstruction; microinstruction format · CPC title