Image Recognition System Using Random Unitary Matrix Encryption
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Facial recognition systems utilizing edge cloud computing face challenges in ensuring high security to prevent data leakage and insufficient utilization of multi-device diversity for improved recognition accuracy.
Innovation Solution
An image recognition system employing a random unitary matrix for end-to-end security and ensemble learning across devices to enhance recognition accuracy, where terminals encrypt images using a random unitary matrix, transfer them to servers for downsampling and processing, and the image recognition server uses encrypted dictionaries to estimate image classes through Orthogonal Matching Pursuit and ensemble learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If edge cloud computing is utilized for facial recognition, then computational efficiency is improved, but data security and privacy protection deteriorate
Solution Approach 1:
The system performs preliminary encryption of facial images at the terminal device before transmission to the edge cloud server. The encryption using random unitary matrices is executed in advance, ensuring that only encrypted data traverses the network and is processed on the edge server, thereby maintaining security while enabling efficient cloud-based computation
Solution Approach 2:
Random unitary matrices serve as an intermediary mechanism between the terminal device and edge cloud server. This mathematical transformation layer allows the system to process encrypted representations of facial images without exposing the actual biometric data, enabling secure computation through an intermediate encrypted domain
2Measurement precision
If centralized cloud processing is used, then recognition accuracy can be improved through aggregated data, but data leakage risks increase
Solution Approach 1:
Encryption is performed preliminarily at the terminal device before data leaves the user's control. This preliminary protective action ensures that even when data is aggregated across multiple devices for improved recognition accuracy, the underlying biometric information remains protected from potential leakage
Solution Approach 2:
The system processes encrypted copies or transformed representations of facial images rather than the original biometric data. The random unitary matrix transformation creates a mathematical representation that preserves recognition capabilities while being unusable for reconstructing the original image, enabling accurate processing without exposure of sensitive data
3Measurement precision
If multiple devices are utilized for ensemble learning, then recognition accuracy is improved, but system complexity increases
Solution Approach 1:
The system segments the ensemble learning process into independent terminal devices, where each device independently generates encrypted representations using the same random unitary matrix framework. This segmentation allows parallel processing across devices while maintaining a unified security protocol, reducing the coordination complexity that would otherwise arise from multi-device collaboration
Solution Approach 2:
The system changes the parameter space by operating in the encrypted domain rather than the original image domain. By transforming the problem into solving for sparse coefficients in an encrypted dictionary, the system enables multiple devices to contribute to ensemble learning through mathematical operations on encrypted data, avoiding the need for complex coordination of raw image processing across devices
Data Source
AI summary
An object of the present invention is to provide an image recognition system, an image recognition server, and an image recognition method having a new high security framework that can achieve utilization of multi-device diversity. The image recognition system according to the present disclosure includes a computationally non-intensive encryption algorithm based on random unitary transformation and achieves a high level of security. In addition, the image recognition system achieves high recognition performance by using ensemble learning to integrate recognition results based on the dictionaries of different devices.