Encrypted Image AI Inference Using CinS Encoding and DFT

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing technologies face challenges in efficiently performing computations on image data using homomorphic encryption, leading to excessive operational burden and potential privacy risks during transmission and processing.

Innovation Solution

An electronic device and server device system that employs CinS encoding and homomorphic encryption to process image data, allowing efficient computation on encrypted data using AI models, reducing operational burden and ensuring privacy by converting images into one-dimensional vectors and performing discrete Fourier transforms before encryption, and utilizing depth-wise convolutional computations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If homomorphic encryption is used to protect image data privacy during transmission and processing, then data security is improved, but computational overhead and operational burden increase significantly

Engineering Contradiction:

Improvedata securityVSAvoidcomputational overhead

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments image data into multiple slots and processes them in parallel using batch operations. By dividing the image into grid portions and encoding each portion into slots that can be processed simultaneously, the system reduces the sequential computational burden of homomorphic encryption while maintaining data security throughout the processing pipeline.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent transforms image data from spatial domain to frequency domain using Fast Fourier Transform (FFT), and then applies CinS encoding to convert the transformed data into a format optimized for homomorphic encryption. This parameter transformation reduces the computational complexity of subsequent encrypted operations by working with frequency-domain representations rather than raw pixel data.

Inventive Principle:

Principle #35Parameter changes

2Ease of manufacture

If traditional encoding methods are used for image data before homomorphic encryption, then encoding simplicity is maintained, but computational efficiency and processing speed deteriorate

Engineering Contradiction:

Improveencoding simplicityVSAvoidprocessing speed

Core Design Contradiction:

Ease of manufactureVSProductivity

Solution Approach 1:

The patent performs Fast Fourier Transform (FFT) and CinS encoding on image data before applying homomorphic encryption. By preprocessing the image data into frequency-domain representations and organizing it into CinS-encoded slots in advance, the system optimizes the subsequent encrypted processing operations, enabling faster batch computations and reducing the number of required bootstrapping levels.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If multiple bootstrapping levels and rotations are performed to maintain precision in homomorphic encryption, then computation accuracy is improved, but operational overhead and processing time increase

Engineering Contradiction:

Improvecomputation accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies CinS encoding to transform image data into a frequency-domain representation that is better suited for homomorphic encryption operations. This encoding method, combined with processing images in batches and using depth-wise separable convolutions, reduces the noise accumulation rate during encrypted computations, thereby reducing the frequency of required bootstrapping operations and decreasing overall processing time while maintaining accuracy.

Inventive Principle:

Principle #35Parameter changes

4Productivity

If image data is transmitted in plaintext to server devices for AI processing, then processing efficiency is improved, but privacy risks and data leakage possibilities increase

Engineering Contradiction:

Improveprocessing efficiencyVSAvoidprivacy risks

Core Design Contradiction:

ProductivityVSObject-affected harmful factors

Solution Approach 1:

The patent introduces CinS-encoded homomorphic ciphertext as an intermediary form between plaintext image data and AI model processing. The image data is encrypted using homomorphic encryption after CinS encoding, allowing the server to perform AI computations on the encrypted data without ever accessing the plaintext. This intermediary encrypted representation enables secure outsourcing of image processing while maintaining computational efficiency through batch operations and optimized encrypted convolution algorithms.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20260012327A1Electronic device, server device for handling homomorphic encrypted data, and methods thereof

Publication Date: 2026.01.08 CRYPTO LAB INC

AI summary

Disclosed is an electronic device. The electronic device includes: a communication unit configured to perform communication with a server device storing an artificial intelligence (AI) model that performs computation in an encrypted state; a memory configured to store a plurality of images; a processor configured to divide each of the plurality of images into a plurality of divided images and list all the divided images in a form of a one-dimensional vector to perform slot encoding, and configured to obtain CinS-encoded data in a form of a frequency signal by performing discrete Fourier transform (DFT) on slot-encoded data in units of slots corresponding to each image and transmit a homomorphic ciphertext obtained by performing homomorphic encryption on the CinS-encoded data to the server device through the communication unit; and a display.