Mean value estimation method and device based on classification transformation disturbance mechanism

Abstract

The invention discloses a mean value estimation method and device based on a classification transformation disturbance mechanism, belongs to the technical field of information security, and adopts a data transformation disturbance mode to divide numeric data into transformation ranges and segment the numeric data, and transform the numeric data into one-dimensional binary classification data according to the segments; the converted data is distributed by using a random response mechanism, and a numerical value is randomly and uniformly extracted from a numerical value section of a disturbed data identifier as a disturbance value; compared with other methods, the method has the advantages that a local differential privacy mechanism is met, meanwhile, high data utility can be obtained in data analysis tasks such as mean value estimation, the classification accuracy of the obtained model is higher, and the performance is better.

Description

technical field [0001] The invention belongs to the technical field of information security, and in particular relates to a mean value estimation method and device based on a classification transformation disturbance mechanism. Background technique [0002] As a branch of differential privacy, the local differential privacy mechanism provides a stronger privacy guarantee than differential privacy. The most typical disturbance mechanism is the random response mechanism. In local differential privacy, assuming that the server is untrustworthy, the user does not directly send the original data to the server, but perturbs the data locally to satisfy local differential privacy, and then sends the perturbed data to the server. The server performs corresponding data analysis tasks on the collected noise data to obtain the required statistical information. Using local differential privacy for privacy protection does not require a large number of complex calculations, and can effect...

AI Technical Summary

Problems solved by technology

However, the accuracy of existing local differential privacy mechanisms is poor, and there is still room for improvement

And most of these methods perturb the data directly, ignoring the possibility of converting the data type for perturbation, ignoring the possibility of perturbing by changing the data type to balance the utility and privacy of the data

Images