"安全的两方协议：基于差分隐私的隐私保护分类器"

This paper discusses the use of differential privacy in privacy protection classification through a secure two-party protocol. The focus of the research is on privacy protection in distributed data mining, where multiple data owners contribute to the mining process in a non-secure environment. The protocol ensures that sensitive information is not disclosed to the data owners, thus preserving individual privacy, while also producing useful mining models. The proposed secure two-party protocol leverages differential privacy to construct privacy-protected decision tree classifiers on distributed data. The protocol employs secure multiparty computation to guarantee privacy protection, and the algorithm utilizes both parallel and sequential combinations. Additionally, the distributed exponential mechanism is applied to ensure that the output is differentially private. The implementation of the protocol on real-world distributed data in a distributed environment yielded promising results. Experimental findings indicate that the decision tree classifiers produced by the protocol are highly practical, while also exhibiting reasonable efficiency and scalability. The thesis, titled "Secure Two-Party Protocol for Privacy-Preserving Classification via Differential Privacy," was submitted by Manish Kumar in December 2019 as partial fulfillment for the degree of Master of Science in Computer Science at Boise State University. The research findings are considered to be an important contribution to the field of privacy protection in distributed data mining, and it provides valuable insights into the development of secure protocols for privacy-preserving classification.