Abstract:

When the current algorithm is used to filter privacy-sensitive data， only a single approximation acquisition method is used. There are certain limitations in obtaining the approximation， which leads to the problem of high MAE and RMSE values. A privacy-sensitive data filtering algorithm based on fuzzy approximation is proposed. First， the data is reduced in dimensionality through the improved local sensitive hash algorithm E²LSH， and low-dimensional data that is more conducive to the subsequent approximation calculation is obtained， and then the Paillier homomorphic encryption algorithm is used to protect the privacy-sensitive data under the premise of ensuring data security. After extraction， the trapezoidal fuzzy scoring model is finally constructed， and the fuzzy approximation is calculated by the mixed model similarity algorithm of modified cosine similarity and Pearson correlation similarity to complete the filtering of privacy-sensitive data. Analysis of the experimental results shows that the minimum MAE value of the proposed method is lower than 0.82， indicating that the method can effectively reduce the MAE value and RMSE value and improve the data filtering effect.

Key words: fuzzy approximation, privacy-sensitive data, data filtering, Paillier homomorphic encryption algorithm, hybrid model algorithm, cosine similarity, Pearson correlation similarity, trapezoidal fuzzy scoring model