Decision tree classification service system and method supporting privacy protection

Abstract

The invention belongs to the field of machine learning and privacy protection, and particularly relates to a decision tree classification service system and method supporting privacy protection. The service system comprises a model owner module, a client module, a cloud service module and a ciphertext operation module. The method comprises a preparation stage and a classification stage. The invention provides a decision tree classifier supporting privacy protection orienting cloud encryption data, and designs and implements a decision tree classification service system supporting privacy protection. Original data cannot be recovered through encrypted data uploaded by a user, privacy protection in the outsourcing calculation process is guaranteed, large-scale data are outsourced to a third-party server which is high in storage and calculation resource, local infrastructure investment and management of the user are reduced, and then more economic benefits are generated.

Description

technical field [0001] The invention belongs to the field of machine learning and privacy protection, and in particular relates to a decision tree classification service system and method supporting privacy protection. Background technique [0002] At present, there are three main types of privacy protection research methods in the process of data classification: 1) Data perturbation technology, which directly perturbs the value of each data record by adding random noise, so that the distribution of perturbed data looks very different from the distribution of actual data; different. However, the perturbed data does not have semantic security, and for the classifier, it cannot produce accurate classification results; 2) Secure Multi-Party Computation (Secure Multi-Party Computation, SMC) and its derivative technologies, such as Security Information Retrieval (PIR) technology , Security Data Mining (PPDM) technology. Such techniques assume that the data set is split horizont...

AI Technical Summary

Problems solved by technology

[0006] In terms of security: the existing research on constructing classifiers does not fully consider the privacy and security protection of data, and the existing schemes can only support the privacy security of one or several of the training data, classification model, user input data and output results , cannot guarantee the privacy and security of all data at the same time

At the same time, there is no effective management of the private key of homomorphic encryption, and the private key data is easily stolen by criminals

[0007] In terms of efficiency: the existing security protocol for constructing classifiers is based on an interactive environment, which has a certain dependence on the network bandwidth of the device. Due to secure multi-party computing, the calculation amount of the participants is the same, so there is also a certain degree of computing power for the device. Requirements are unacceptable for some lightweight devices with limited resources. At the same time, the execution of the protocol requires multiple interactions and a large amount of data transmission, and network delay will also affect the execution of the protocol.

[0008] In terms of usability: in the existing research on constructing classifiers, the structure of classifiers is too simple, and the existing schemes only design security protocols for specific classifiers, which lack universality and cannot adapt to a wide range of practical application scenarios

At the same time, during the classification process of the machine learning classifier, the model provider and the user are required to be online all the way, and the user still needs to participate in a large number of ciphertext calculations, which is contrary to the original intention of the user delivery service provider for data classification prediction and lacks practicability

Images