ITR: Private Prediction Using Selective Models

本研究将研究SPMP（Selective Private Model-basedPrediction）问题，该问题在以下情况下被证明：具有私有输入的客户端希望使用服务器的私有模型进行预测;然而，双方都不希望将其私有数据透露给任何人。更具体地说，本研究将研究的问题，为选定的模型，包括隐马尔可夫模型，神经网络模型，贝叶斯网络模型，决策树模型。我们的目标是开发有效和实用的解决方案，使这种类型的隐私保护预测。该项目将研究两种方法：商品服务器的方法和多服务器的方法。商品服务器方法使用来自第三方的数据来保护私人数据的机密性。多服务器方法使用重复的服务器，因此单个服务器无法了解客户端数据的所有信息。基于这两种方法，我们还将研究各种数据伪装技术，有效地解决SPMP问题，使模型所有者能够提供新形式的电子商务服务，同时保护客户的隐私信息。此外，从拟议的活动中获得的结果，方法和积木可以提供对安全多方计算（SMC）研究的宝贵理解和见解，并有助于推进和扩展SMC研究的领域。

This research will study the SPMP (Selective Private Model-basedPrediction) problem demonstrated with the following scenario: A clientwith a private input wants to use a server's private model to makepredictions; however, neither side wants to disclose its private data toanybody. More specifically, the research will study the problem forselected models including the Hidden Markov Model, the Neural NetworkModel, the Bayesian Network Model, and the Decision Trees Model. The goalis to develop efficient and practical solutions that enable such type ofprivacy preserving prediction.The project will investigate two approaches: commodity-server approach andmultiple-server approach. The commodity-server approach uses thecommodities (data) from a third party to preserve the confidentiality ofthe private data. The multiple-server approach uses duplicate servers sono single server can learn all information about the client's data. Basedon these two approaches, various data disguising techniques will also bestudied.Efficient solutions to the SPMP problems enable model owners to providenew forms of e-commerce services while protecting customers' privateinformation. Furthermore, the results, methodologies, and the buildingblocks gained from the proposed activity can provide invaluableunderstanding and insights into the Secure Multiparty Computation (SMC)research, and help to advance and expand the areas of SMC research.---------------------------------------