AI Server Feature Reconstruction for Privacy-Safe Model Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI systems face privacy violations when transferring raw sensing data, such as images or voice recordings, to servers for recognition model updates, as they may contain sensitive or private information.
Innovation Solution
An AI server that receives feature data extracted from sensing data and uses deep learning models to generate second sensing data, which is similar but anonymized, for updating recognition models, thereby preserving user privacy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If raw sensing data is transmitted to the AI server for recognition model updates, then the model performance can be improved, but user privacy is violated
Solution Approach 1:
The patent extracts only the essential feature data from raw sensing data before transmission. The AI device performs feature extraction locally, identifying and transmitting only the critical characteristics needed for model improvement while leaving out personally identifiable information and sensitive content, thus resolving the contradiction between model performance and privacy protection
Solution Approach 2:
The patent introduces feature data as an intermediary between raw sensing data and the recognition model. This intermediate representation preserves the essential information needed for model training while removing direct privacy risks, acting as a mediator that satisfies both model improvement requirements and privacy protection concerns
2Object-affected harmful factors
If feature data is extracted and transmitted instead of raw sensing data, then privacy is protected, but the amount of useful information for model training is reduced
Solution Approach 1:
The patent applies local quality by preserving different levels of detail in different parts of the data. Essential features needed for model training are maintained with high fidelity, while personally identifiable information and sensitive content are removed or generalized, creating a non-uniform data structure that optimizes both privacy protection and training utility
Solution Approach 2:
The patent transforms raw sensing data into feature data by changing the data parameters and representation format. This transformation process converts detailed raw data into extracted features that maintain the statistical and structural properties necessary for model improvement while eliminating privacy risks associated with the original data format
Data Source
AI summary
Disclosed is an artificial intelligence (AI) server. The AI server includes a communication unit configured to communicate with an AI device; and an AI unit configured to receive feature data from the AI device, wherein the received feature data is generated by the AI device by obtaining sensing data and compressing the sensing data while preserving a feature of the sensing data; and input the received feature data to a deep learning model to obtain second sensing data for use in a recognition model related to an AI function of the AI device.