Multi-layer differential privacy embedded decision tree model-based privacy risk control method

Abstract

The invention provides a multi-layer differential privacy embedded decision tree model-based privacy risk control method. The method includes the following steps that: initialization is carried out; a differential privacy technology is embedded into a multi-layer decision tree model; and a multi-layer decision tree is obtained. According to the method of the invention, a multi-layer embedding mode is adopted to embed the differential privacy technology into the decision tree model. Compared with the prior art, the decision tree model can be under the protection of differential privacies, and the prediction accuracy of the model is greatly improved.

Description

technical field [0001] The invention relates to a privacy protection technology differential privacy and a data mining technology decision tree model, in particular to a privacy risk control method for embedding multi-layer differential privacy into a decision tree model. Background technique [0002] Differential privacy technology is a new privacy protection technology that can quantitatively control the risk of privacy leakage. Embedding differential privacy into the decision tree model can protect the decision tree model. In the previous research, the method of embedding differential privacy into the decision tree model is a layer of embedding. [0003] The realization of differential privacy requires the implementation of exponential mechanism and Laplacian mechanism, which contains uncertainty and noise, so the prediction accuracy of the decision tree model with differential privacy will be very low. It is lower than the tolerance of error in industrial systems, so t...

AI Technical Summary

Problems solved by technology

[0003] The realization of differential privacy requires the implementation of exponential mechanism and Laplacian mechanism, which contains uncertainty and noise, so the prediction accuracy of the decision tree model with differential privacy will be very low

It is lower than the tolerance of error in industrial systems, so the decision tree model with differential privacy protection has not been applied in industrial systems

[0004] Using this algorithm can realize a decision tree model with differential privacy technology protection, but the prediction accuracy of the decision tree model of this algorithm is low and cannot be applied to industrial systems

Images