Outsourced privacy-preserving decision tree classification service over encrypted data. (August 2020)

• Record Type: Journal Article
• Authors: Wang, Chen; Wang, Andi; Xu, Jian; Wang, Qiang; Zhou, Fucai
• Is Part Of: Journal of information security and applications. Volume 53(2020)
• Publication Date: 2020-08
• Subjects: Privacy preserving -- Machine learning -- Decision tree -- Cloud service; Computer security -- Periodicals; Information technology -- Security measures -- Periodicals; 005.805
• DOI: 10.1016/j.jisa.2020.102517
• Languages: English
• ISSNs: 2214-2126

Abstract

Abstract: The classification is one of the important machine learning technologies, which plays an important role in the fields of medical treatment, image processing, and cyberspace security. In recent years, the popularity of cloud computing has caused many data owners to seek outsourced storage computing, which has become a new trend in the development of classification. Although it can be used in several research areas, such as medical and financial, this paradigm introduces new security and privacy problems. In response to the ciphertext classification requirements in different scenarios in outsourcing environments, this paper proposed an outsourced privacy-preserving decision tree classification model based on homomorphic encryption, which can hide sensitive inputs (model and query) and outputs (analytic result) from a counterparty. A secure exclusive OR (XOR) protocol is created; it intelligently computes the XOR of two bit-encrypted data to obtain another bit-encrypted result. The experimental results show that the proposed protocols are particularly efficient for outsourced decision trees, which are encrypted and stored in the cloud. These decision trees are typical of classification models trained from real datasets, ranging from heart disease to cancer classification.