A new data clustering strategy for enhancing mutual privacy in healthcare IoT systems

Abstract In the rapidly growing era of Internet-of-Things (IoT), healthcare systems have enabled a sea of connections of physical sensors. Data analysis methods (e.g., k-means) are often used to process data collected from wireless sensor networks to provide treatment advices for physicians and patients. However, many methods pose a threat of privacy leakage during the process of data handling. To address privacy issues, we propose a mutual privacy-preserving k-means strategy (M-PPKS) based on homomorphic encryption that neither discloses the participant’s privacy nor leaks the cluster center’s private data. The proposed M-PPKS divides each iteration of a k-means algorithm into two stages: finding the nearest cluster center for each participant, followed by computing a new center for each cluster. In both phases, the cluster center is confidential to participants, and the private information of each participant is not accessible by an analyst. Besides, M-PPKS introduces a third-party cloud platform to reduce the communication complexity of homomorphic encryption. Extensive privacy analysis and performance evaluation results manifest that the proposed M-PPKS strategy can achieve high performance. In addition, it can obtain approximate clustering results efficiently while preserving mutual private data.