Abstract
Due to the increasing healthcare costs and the advance of wireless technology, health monitoring systems have been widely adopted recently. In health monitoring systems, a hospital outsources a clinical decision model to a cloud service provider, which receives biomedical data from remote clients and produces clinical decisions based on the outsourced model. Due to critical privacy concerns, both the clinical decision model and biomedical data should be protected. In this article, we propose an efficient and privacy-preserving decision tree (PPDT) classification scheme for health monitoring systems. Specifically, we first transform a decision tree classifier (i.e., the clinical decision model) into the Boolean vectors. Then, we leverage symmetric key encryption to encrypt the Boolean vectors as encrypted indices. The PPDT classification is achieved by searching the encrypted indices with encrypted tokens. We formulate a leakage function and provide the security definition and simulation-based proof for PPDT. The performance analyses demonstrate that PPDT is very efficient in terms of computation, communication, and storage. Experimental evaluations show that PPDT only requires microsecond-level execution time, kilobyte-level communication costs, and kilobyte-level storage costs on the test data set.
Funding Information
  • National Natural Science Foundation of China (U20A20174, 61772191, 61902123, 62002112)
  • China Scholarship Council (201806130132)
  • National Key Research and Development Projects (2018YFB0704000)
  • Science and Technology Key Projects of Hunan Province (2015TP1004, 2018TP2023, 2019WK2072)
  • Natural Sciences and Engineering Research Council (NSERC) of Canada
  • China Postdoctoral Science Foundation (2020M672488)
  • Natural Science Foundation of Hunan Province (2020JJ5085)
  • Science and Technology Key Projects of Changsha City (kq2004025, kq2004027, kq2006029)

This publication has 36 references indexed in Scilit: