IoT Data Model Training with Noise for Privacy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The Internet of Things (IoT) faces significant challenges in protecting personal data privacy due to inadequate security measures, with traditional centralized deep learning methods risking misuse of user data.
Innovation Solution
A privacy-protection-based data processing model acquisition method and apparatus that acquires sensor data from IoT devices, trains initial models, integrates them, processes new data with random noise to determine label categories, and trains integrated models to construct machine learning models without sharing data, ensuring better performance and enhanced privacy protection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional centralized deep learning is used, then model training can be performed uniformly by cloud servers, but user data privacy is compromised due to data submission to data centers
Solution Approach 1:
The patent segments the centralized training process into distributed local training. Each participant trains a local model using their own data independently, then only the model parameters (not the data itself) are shared and aggregated. This segmentation eliminates the need to submit raw data to centralized servers while maintaining training efficiency through parallel processing.
Solution Approach 2:
The patent introduces an intermediary aggregation server that collects and aggregates model parameters from multiple participants without accessing the underlying data. This intermediary acts as a mediator between participants and the central system, enabling coordinated training while preserving data privacy by never exposing raw data to centralized locations.
2Measurement precision
If data is shared among multiple participants to improve model performance, then model accuracy increases, but privacy protection deteriorates
Solution Approach 1:
The patent extracts only the essential training information (model parameters) from the data while leaving the data itself private. By taking out only the parameters and not the raw data, the system enables model accuracy improvement through parameter aggregation while preventing privacy leakage since the actual data never leaves local devices.
Solution Approach 2:
The patent creates copies of model parameters from each participant's local model and aggregates these copies. This copying approach allows the system to combine information from multiple participants to improve accuracy while the original data remains unchanged and private at each participant's location.
3Object-affected harmful factors
If noise is introduced into the model processing, then privacy protection is enhanced by covering individual contributions, but processing complexity increases
Solution Approach 1:
The patent converts the potentially harmful effect of noise (which could degrade model performance) into a beneficial privacy protection mechanism. By intentionally introducing noise that obscures individual contributions, the system achieves privacy protection while the overall training process remains manageable through automated parameter aggregation.
Data Source
AI summary
Disclosed are a privacy-protection-based data processing model acquisition method and apparatus, a terminal device and a storage medium. The method includes: acquiring sensor data of a plurality of sensors of a preset Internet of Things device; training an initial data model corresponding to each of the sensors through the sensor data corresponding to the sensor to obtain an intermediate data model corresponding to each of the sensors, and integrating the intermediate data models corresponding to the sensors to form an integrated data model; processing new data through the integrated data model and random noise to acquire a label category corresponding to the new data; and training the integrated data model according to the new data and the label category of the new data to acquire a data model. The method solved the technical problem of poor privacy protection of Internet of Things data.