Encrypted Face Vector Matching Without Decryption Delay
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing communication systems face challenges in maintaining security and efficiency during encrypted data operations, particularly in identifying users using biometric information, where decryption processes consume resources and time, and there is a risk of data leakage.
Innovation Solution
A server system that utilizes homomorphic encryption to transform and encrypt feature vectors, allowing for secure and efficient user identification through inner product operations without decrypting the data, using a transformation that converts feature vector components into 1, −1, or 0 based on their magnitudes and signs, and performs operations on encrypted databases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If decryption is performed to use encrypted messages, then the messages can be accessed and used, but resources and time are wasted in the decryption process
Solution Approach 1:
The patent replaces the traditional decryption mechanism with homomorphic encryption operations. Instead of decrypting messages to perform operations (mechanical decryption process), the system performs computations directly on encrypted data using homomorphic properties, substituting the decryption step with encryption-compatible operations that yield the same functional result.
2Ease of operation
If decryption is performed to use encrypted messages, then the messages can be accessed, but security risk increases as messages may be leaked to third parties
Solution Approach 1:
The patent eliminates the decryption step entirely by using homomorphic encryption operations that work directly on encrypted data. This substitution removes the security vulnerability point where decrypted messages could be exposed, while still allowing legitimate operations to be performed on the data.
Solution Approach 2:
The patent introduces homomorphic encryption as an intermediary mechanism that allows operations on encrypted data without exposing the plaintext. This intermediary layer enables computation while maintaining security, acting as a bridge between data usage requirements and security preservation.
3Productivity
If feature vectors are transformed into binary vectors for homomorphic encryption, then computational speed improves, but accuracy may be reduced
Solution Approach 1:
The patent transforms feature vectors from continuous floating-point values to discrete binary vectors through thresholding operations. This parameter change from continuous to discrete domain enables efficient homomorphic encryption operations while preserving the essential discriminative information needed for accurate user identification.
Data Source
AI summary
Disclosed is a server. The server includes a camera, a memory configured to store an encrypted database generated based on a plurality of feature vectors corresponding to a plurality of users, and a processor configured to acquire an image obtained by capturing a target user's face using the camera and transform a feature vector acquired from the image based on transformation and identify the target user based on the transformed feature vector and the encrypted database. The encrypted database is generated by transforming the plurality of feature vectors based on the transformation and homomorphically encrypting the plurality of transformed feature vectors. The transformation transforms components of the feature vectors into 1, −1, or 0 depending on magnitudes and signs of the components of the feature vectors.