Homomorphic Ciphertext Matrix Splitting for Secure AI Computation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication technologies face inefficiencies in performing high-dimensional matrix operations with homomorphic encryption, leading to resource waste and potential data leakage during decryption.
Innovation Solution
A method for processing homomorphic ciphertexts involves splitting a matrix ciphertext into two parts, ensuring identical a-parts and differing b-parts, and performing operations within an error range, while converting modulus and ring degrees to efficiently perform matrix operations without decryption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If homomorphic encryption is used to perform operations on encrypted data, then data security is improved, but computational complexity and resource consumption increase
Solution Approach 1:
The patent applies segmentation by dividing the homomorphic ciphertext into multiple parts (first ciphertext part and second ciphertext part) before performing matrix operations. This segmentation allows the system to process encrypted data in manageable segments, reducing the overall computational burden while maintaining security. The first and second ciphertext parts are generated by splitting the original ciphertext, enabling efficient parallel or sequential processing that lowers resource consumption compared to operating on the complete ciphertext as a single unit.
2Ease of operation
If decryption is performed to use encrypted messages, then operational flexibility is improved, but data leakage risk and resource waste increase
Solution Approach 1:
The patent introduces an intermediary approach by performing matrix operations directly on the homomorphic ciphertext and its parts without complete decryption. The system uses the ciphertext itself as the operating medium, applying transformations and computations in the encrypted domain. This intermediary processing allows operational flexibility to be maintained while eliminating the need to fully decrypt data, thereby preventing data leakage to unauthorized parties including the processing system itself.
3Adaptability or versatility
If high-dimensional matrix operations are performed on homomorphic ciphertexts, then functionality for AI applications is improved, but computational efficiency deteriorates
Solution Approach 1:
The patent applies segmentation by dividing the high-dimensional matrix operation into operations on smaller ciphertext parts. Instead of performing computationally intensive matrix multiplication on the entire large ciphertext, the system operates on segmented portions (first and second ciphertext parts), significantly improving computational efficiency while maintaining the ability to perform AI/ML inference tasks.
Solution Approach 2:
The patent implements partial action by performing matrix operations only on necessary portions of the ciphertext rather than the complete dataset. The system selectively processes the first and second ciphertext parts that are relevant to the specific AI application task, avoiding unnecessary computations on the entire encrypted dataset, thereby improving productivity without sacrificing the required AI functionality.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Provided is an electronic apparatus including: a memory for storing an instruction; and a processor configured to execute the instruction, wherein the processor is configured to generate a first matrix ciphertext by disposing a plurality of homomorphic ciphertexts in a matrix form if a matrix operation command for the plurality of homomorphic ciphertexts is input, split the first matrix ciphertext into a second matrix ciphertext and a third matrix ciphertext that satisfy predetermined conditions, and acquire a matrix operation result by performing a matrix operation between each of the second matrix ciphertext and the third matrix ciphertext and a plaintext matrix corresponding to the matrix operation command.