基于云平台的密文数据离群点检测算法

Outlier Detection Algorithm for Cryptographic Data Based on Cloud Platform

摘要: 将数据和计算外包给云服务器为大规模数据存储和查询处理是一种经济高效的方法。但是由于安全和隐私问题，我们需要在使用云服务器计算过程中保护诸如医疗记录这种敏感数据。一种方法是将加密过的数据外包给云服务器，并让云服务器仅对加密数据执行查询处理，这样云服务器就不会获得有关数据、查询和查询结果的任何知识。在不进行数据云解密的情况下进行加密数据的查询任务是非常具有挑战性的。我们设计的任务是在加密数据条件下云服务器进行安全K近邻(KNN)查询方法进行离群点检测，即用户发送一条加密查询给具有加密数据集的云服务器，并得到该数据是否为离群点的信息。我们首先提出了一个简单方案，并证明该方案并非是完全安全的。之后我们提出了具有更高安全性的安全方案，它能够有效保护用户查询数据、服务器数据集和数据访问模式的机密性。

Abstract: Outsourcing data and computing to cloud servers for large-scale data storage and query processing is an economical and efficient method. However, due to security and privacy issues, we need to protect sensitive data such as medical records in the process of using cloud server computing. One ap-proach is to outsource the encrypted data to a cloud server and have the cloud server perform only query processing on the encrypted data, so that the cloud server does not gain any knowledge of the data, queries, and query results. The query task of encrypting data without data cloud decryption is very challenging. The task we designed is to detect outliers with the cloud server’s secure k-nearest neighbor (KNN) query method under the condition of encrypted data, that is, the user sends an encrypted query to the cloud server with encrypted data set and obtains the information whether the data is an outlier or not. We first propose a simple scheme and prove that the scheme is not completely safe. Then we proposed a security scheme with higher security, which can effectively protect the confidentiality of user query data, server data set and data access mode.