Design and Security Analysis of Cryptographic Protocol for Privacy-Preserving Data Mining

在数据挖掘过程中，由于需要处理健康数据、个人偏好等，存在着隐私泄露的问题。有一些保护隐私的数据挖掘方法，在原始数据中加入随机噪声数据，或者在不破坏原始数据结构的情况下进行计算和通信。在本研究中，我们设计了可以从理论上证明其安全性的加密协议。对于安全模型，我们使用通用可组合性（UC），其中可以安全地组合多个协议。我们设计了仅在以前的安全模型中证明了安全性的数据挖掘协议，如遗忘传输协议，并证明了UC概念中的安全性。

In the process of data mining, there exists a problem about privacy breach because we need to deal with health data, personal preferences, etc. There are approaches to privacy preserving data mining in which random noisy data are added to the original data or computation and communication are performed without breaking the structure of original data. In this research, we designed cryptographic protocols whose security can be proved in a theoretical way. For the security model, we use Universal Composability (UC) in which combining several protocols can be done securely. We designed protocols such as oblivious transfer for data mining whose security was proved only in the previous security model, and proved the security in the notion of UC.