Abstract:Rule–based fuzzy models have modular architectures and come with well-developed design methodologies such that they can build accurate models with good interpretabilities in system modeling. However, a large amount of private data needs to be used for statistical analysis and forecasting in rule–based fuzzy models. The purpose of this study is to build an intelligent model with high accuracy and versatility under the premise of data privacy and model security. To mitigate the risk of malicious attacks on privacy data during the analysis process, we have employed highly regarded differential privacy techniques to devise a novel rule-based fuzzy modeling approach. We propose a function approximation mechanism to reconstruct the objective function and add a perturbation mechanism to the objective function in Takagi-Sugeno rules model. Taking into account the delicate balance between data privacy and utility, we have innovatively introduced a Takagi-Sugeno rule-based model based on differential privacy. This model is applicable to both linear and nonlinear systems, offering protection to sensitive data privacy and model security within the system. We investigate the relationship between the interpretability of the model and the degree of privacy protection. By constructing a reasonable rule base, we achieve higher accuracy than other system modeling methods based on differential privacy. This paper compares the influence of the number of rules on differential privacy, and considers the algorithm performance under various noise distributions. It is shown that the Takagi-Sugeno rules model based on differential privacy has a strong ability to predict and analyze data.

Fuzzy Prediction Model in Privacy Protection: Takagi-Sugeno Rules Model Via Differential Privacy

Protecting Regression Models with Personalized Local Differential Privacy

Functional Mechanism: Regression Analysis under Differential Privacy

Differentially Private and Fair Classification Via Calibrated Functional Mechanism.

Differential Privacy Preservation in Regression Analysis Based on Relevance

A Practical Differentially Private Support Vector Machine.

A Privacy Protection Model of Data Publication Based on Game Theory

Recommender Systems Based On Autoencoder And Differential Privacy

Research on Recommender System Model Based on Differential Privacy and Time Series

Deep Learning with Gaussian Differential Privacy

Differentially-Private Deep Learning from an Optimization Perspective.

Differentially private regression analysis with dynamic privacy allocation

More Than Privacy: Applying Differential Privacy in Key Areas of Artificial Intelligence

Research on the Application of Deep Learning-based Differential Privacy Protection Models in Financial Big Data

Differential Privacy for Class-based Data: A Practical Gaussian Mechanism

Privacy-Preserving Collaborative Web Services QoS Prediction Via Differential Privacy.

One Parameter Defense -- Defending against Data Inference Attacks via Differential Privacy

Differential privacy in probabilistic systems.

Model Explanations with Differential Privacy

Differential Privacy for Credit Risk Model

Statistical Inference in the Differential Privacy Model