A quantitative evaluation method and system for privacy protection in a multi-party data collaboration scenario

Abstract

The present invention discloses a privacy protection quantitative evaluation method and system in a multi-party data collaboration scenario. In the method, the data resource provider inputs the data resource into a model for training according to an algorithm and a protection mechanism, and the output satisfies the model index and threshold At the same time, the data resource provider uses attack means to attack the above model, records and quantifies the privacy leakage of the model, and outputs the privacy index; the data resource user and the data resource provider respectively meet their own needs according to the model index and privacy index , to make a decision on whether to carry out data cooperation. The invention improves the effective assessment and information disclosure of privacy risk and data value in the process of data sharing, effectively solves the problem of information asymmetry between the data resource provider and the data use demander, and helps to build a healthier and more sustainable Developed data sharing, interaction and transaction system.

Description

Technical field [0001] The present invention belongs to the field of network information, and in particular to a privacy protection quantization assessment method and system under multi-data collaborative scenarios. Background technique [0002] In the future, digital and intelligent development in the fields of finance, medical, transportation will pay more attention to the joint modeling of cross-institutional data sharing and distributed artificial intelligence algorithms. This type of application involves distributed data collection, transmission, storage, use, and across mechanism data sharing, joint modeling, and system models complicated. Relevant privacy protection algorithms involve comprehensive use of multi-class technologies such as artificial intelligence, data encryption, and network security. In recent years, although research on the application scenario has initially produced a number of solutions, such as the distributed artificial intelligence algorithm framewor...

AI Technical Summary

Problems solved by technology

Such frameworks mostly focus on solving a series of engineering problems such as high computational complexity, high communication costs, and integration with existing machine learning and deep learning algorithms, and aim to achieve preliminary proof of concept, but they do not strictly associate various algorithms. The privacy attack model is incorporated into the framework, so it does not have the ability to quantitatively evaluate the complete privacy protection utility and data utility

At present, academia and industry at home and abroad are also trying to build data sharing, interaction and transaction systems based on multi-party data sharing scenarios. Through this type of system, the supply and demand matching between data resource parties and data demand parties can be realized. However, such systems cannot provide privacy protection. Therefore, it is impossible to provide effective privacy disclosure risk warnings to data resource providers, and it is impossible to provide data utility to data users, especially to provide risk warnings on the decline of data utility under privacy protection conditions. Therefore, the Both the data resource provider and the data demander in this type of system lack sufficient auxiliary decision-making basis for data transactions, so the matching effect of the system is limited.

Images