IP Library › Granted Patent US 12,395,316
Granted Patent B2
US 12,395,316 · App. 18/196,687 · Granted Aug 19, 2025

Homomorphic encrypted one-hot maps using intermediate chinese remainder theorem (CRT) maps

Inventors: Nir Drucker (Zichron Yaakov, IL); Ramy Masalha (Kafr Qari, IL); Hayim Shaul (Kafr Saba, IL)
Assignee: International Business Machines Corporation
H04L9/008
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,395,316
App. No.
18/196,687
Granted
Aug 19, 2025
Kind
B2
Abstract

A method, apparatus and computer program product for privacy-preserving homomorphic inferencing using one-hot data representations. In this approach, a client interacting with a cloud-based server submits one-hot maps of a Chinese Remainder Theorem (CRT)-based representation of an data element, and the server expands these maps in an online phase to obtain a full one-hot map for the element. After the server obtains the full one-hot map, it performs an operation, e.g., a comparison operation associated with inferencing on a decision tree, on the one-hot map under homomorphic encryption, and in response generates a result. The result is provided back to the client.

Claims (42)

1. A method implemented in part on at least one processor for secure computations under Homomorphic Encryption (HE), comprising:

receiving an encrypted data point that encrypts one or more elements, wherein at least one element is associated with a plaintext number, the plaintext number being represented in the encrypted data point as a set of one-hot maps, each of the set of one-hot maps corresponding to a coprime factor computed for the plaintext number using a Chinese Remainder Theorem (CRT) encoding;

under homomorphic encryption, generating a one-hot map representation of the plaintext number using the set of one-hot maps;

under homomorphic encryption, performing an operation on the one-hot representation and generating a result; and

providing the result as a response to the encrypted data point.

2. The method as described in claim 1 wherein generating the one-hot map representation comprises:

for a given one-hot map of the set of one-hot maps, duplicating the given one-hot map a number of times to generate a vector, the number of times derived by dividing a result of multiplying the coprime factors by the coprime factor for the given one-hot map; and

multiplying the vectors.

3. The method as described in claim 1 wherein the plaintext number is a number without coprime factors and that has been padded with one or more zeros.

4. The method as described in claim 1 wherein the operation is a comparison operation associated with evaluation of a decision tree.

5. The method as described in claim 1 wherein the homomorphic encryption is based on a Cheon-Kim-Kim-Song (CKKS) scheme.

6. The method as described in claim 1 wherein the one or more elements are arranged according to a tile tensor-based packing scheme.

7. The method as described in claim 1 wherein the CRT encoding is hierarchical, and wherein generating the one-hot map representation is recursively repeated at each hierarchy.

8. An apparatus, comprising:

a processor;

computer memory holding computer program instructions executed by the processor to provide secure computations under Homomorphic Encryption (HE), the computer program instructions comprising program code configured to:

receive an encrypted data point that encrypts one or more elements, wherein at least one element is associated with a plaintext number, the plaintext number being represented in the encrypted data point as a set of one-hot maps, each of the set of one-hot maps corresponding to a coprime factor computed for the plaintext number using a Chinese Remainder Theorem (CRT) encoding;

under homomorphic encryption, generate a one-hot map representation of the plaintext number using the set of one-hot maps;

under homomorphic encryption, perform an operation on the one-hot representation and generating a result; and

provide the result as a response to the encrypted data point.

9. The apparatus as described in claim 8 wherein the program code configured to generate the one-hot map representation comprises program code configured to:

for a given one-hot map of the set of one-hot maps, duplicate the given one-hot map a number of times to generate a vector, the number of times derived by dividing a result of multiplying the coprime factors by the coprime factor for the given one-hot map; and

multiply the vectors.

10. The apparatus as described in claim 8 wherein the plaintext number is a number without coprime factors and that has been padded with one or more zeros.

11. The apparatus as described in claim 8 wherein the operation is a comparison operation associated with evaluation of a decision tree.

12. The apparatus as described in claim 8 wherein the homomorphic encryption is based on a Cheon-Kim-Kim-Song (CKKS) scheme.

13. The apparatus as described in claim 8 wherein the one or more elements are arranged according to a tile tensor-based packing scheme.

14. The apparatus as described in claim 8 wherein the CRT encoding is hierarchical, and the computer program code configured to generate the one-hot map representation is recursively repeated at each hierarchy.

15. A computer program product in a non-transitory computer readable medium, the computer program product holding computer program instructions that, when executed by a first processor in a first processing system and a second processor in a second processing system, provide secure computations under Homomorphic Encryption (HE), the computer program instructions comprising program code configured to:

generate an encrypted data point that encrypts one or more elements, wherein at least one element is associated with a plaintext number, the plaintext number being represented in the encrypted data point as a set of one-hot maps, each of the set of one-hot maps corresponding to a coprime factor computed for the plaintext number using a Chinese Remainder Theorem (CRT) encoding;

output the encrypted data point from the first processing system;

receive the encrypted data point at the second processing system;

under homomorphic encryption, generate a one-hot map representation of the plaintext number using the set of one-hot maps;

under homomorphic encryption, perform an operation on the one-hot representation and generate a result; and

provide the result as a response to the encrypted data point from the second processing system.

16. The computer program product as described in claim 15 wherein the program code configured to generate the one-hot map representation comprises program code configured to:

for a given one-hot map of the set of one-bot maps, duplicate the given one-hot map a number of times to generate a vector, the number of times derived by dividing a result of multiplying the coprime factors by the coprime factor for the given one-hot map; and

multiply the vectors.

17. The computer program product as described in claim 15 wherein the plaintext number is a number without coprime factors and that has been padded with one or more zeros.

18. The computer program product as described in claim 15 wherein the operation is a comparison operation associated with evaluation of a decision tree.

19. The computer program product as described in claim 15 wherein the one or more elements are arranged according to a tile tensor-based packing scheme.

20. The computer program product as described in claim 15 wherein the CRT encoding is hierarchical, and the computer program code configured to generate the one-hot map representation is recursively repeated at each hierarchy.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded May 12, 2023
From: DRUCKER, NIR; MASALHA, RAMY; SHAUL, HAYIM
To: INTERNATIONAL BUSINESS MACHINES CORPORATION
Reel/Frame 063625/0982 →
Continuity (1)
Related Publication 20240380569A1 · Nov 14, 2024
References Cited (19)
US 9436835B1 · Saldamli · 2016 [cited by applicant]
US 10211975B2 · Loftus et al. · 2019 [cited by applicant]
US 20200389303A1 · Kim · 2020 [cited by examiner]
US 20210376995A1 · Ratha · 2021 [cited by applicant]
US 20220224515A1 · Yampolsky · 2022 [cited by examiner]
US 20220321321A1 · Ghosh et al. · 2022 [cited by applicant]
US 20230327856A1 · Lu · 2023 [cited by examiner]
US 20240330500A1 · Zaret · 2024 [cited by examiner]
IN 201102903I1 · 2023 [cited by examiner]
KR 20230076723A · 2023 [cited by examiner]
Zheng, Zhiyong, Fengxia Liu, and Kun Tian. “An unbounded fully homomorphic encryption scheme based on ideal lattices and Chinese remainder theorem.” arXiv preprint arXiv:2301.12060 (2023). (Year: 2023). [cited by examiner]
Kim et al., “CRT-based Fully Homomorphic Encryption over the Integers,”. [cited by applicant]
Aharoni, et al., “HeLayers: A Tile Tensors Framework for Large Neural Networks on Encrypted Data,” arXiv:2011.01805v3 [cs.CR] Jan. 1, 2023. [cited by applicant]
Aslett, et al., “Encrypted statistical machine learning: new privacy preserving methods,” arXiv:1508.06845v1 [stat.ML] Aug. 27, 2015. [cited by applicant]
Brakerski, et al., “Fully homomorphic encryption without bootstrapping,” ACM Transactions on Computation Theory (TOCT), 6(3):1-36, 2014. doi:10.1145/2633600. [cited by applicant]
Cheon, et al., “Homomorphic encryption for arithmetic of approximate numbers,” International Association of Cryptographic Research, 2017. [cited by applicant]
Aharoni et al., “Generating One-Hot Maps under Encryption”, arXiv:2306.06739v1, Jun. 11, 2023, pp. 1-21. [cited by applicant]
Sav et al., “Poseidon: Privacy-Preserving Federated Neural Network Learning”, arXiv:2009.00349, Sep. 1, 2020, pp. 1-24. [cited by applicant]
International Search Report, PCT Application No. IB2024/052671, dated Jun. 20, 2024, 14 pages. [cited by applicant]