Secure Collaborative Outsourced k-Nearest Neighbor Classification with Multiple Owners in Cloud Environment

详细信息    查看全文

• 关键词：Big data ; Cloud computing ; Privacy ; preserving data mining ; K ; nearest neighbor classification ; Multiple owners
• 刊名：Lecture Notes in Computer Science
• 出版年：2017
• 出版时间：2017
• 年：2017
• 卷：10143
• 期：1
• 页码：451-471
• 丛书名：Information Security and Cryptology
• ISBN：978-3-319-54705-3
• 卷排序：10143

文摘

With the advent of big data era, it’s becoming an increasing trend for different clients lack of computational resources to cooperate in outsourcing data mining tasks to cloud service providers in order to produce maximum value of the joint database. Generally, the outsourced data contributed by clients should be encrypted under different keys owing to security concerns. Unfortunately, existing privacy-preserving outsourcing protocols are either restricted to a single key setting or quite inefficient due to frequent server-client interactions, making the deployment far from practical. In this paper, we focus on outsourced k-Nearest Neighbor (kNN) classification over encrypted data under multiple keys, and propose a set of secure building blocks and the Secure Collaborative Outsourced kNN (SCOkNN) protocol. Theoretical analysis shows that the proposed protocol protects the confidentiality of data from data owners, privacy of query, and access patterns in the semi-honest model with negligible computation and communication costs. Experimental evaluation also demonstrates its practicability and efficiency.