K-NEAREST NEIGHBOR CLASSIFICATION OVER SEMANTICALLY SECURE ENCRYPTED RELATIONAL DATA

Abstract

Banking, medicine, scientific research, and government organizations are just a few of the many industries that put data mining to use. Data mining applications make heavy use of the classification process. Recent years have seen a proliferation of theoretical and practical answers to the classification challenge in response to rising privacy concerns. Multiple security models have allowed for the development of these fixes. As cloud computing grows in popularity, more and more people are taking advantage of its encryption capabilities and outsourcing data mining jobs to remote servers. Cloud data encryption makes obsolete the present privacy protection classification systems. An answer to the problem of how to label encrypted information is the driving force behind this paper. In particular, we suggest employing a safe k-NN classifier made for encrypted cloud-based data. The proposed protocol's goal is to keep all information, including user searches and access patterns, private. To our knowledge, our study is the first to employ the semi-honest model in order to create a trustworthy k-NN classifier capable of handling encrypted data. We also perform an empirical test of our suggested protocol using a dataset collected in the wild and a wide range of tuning parameters to see how well it performs.