SSCPDT: A Secure Decision Tree Classification Scheme for Telemedicine Diagnosis

With the development of smart healthcare, telemedicine diagnosis, which provides diagnostic services to users at any time and place, has garnered significant attention. However, the sensitivity of medical data causes it to be an important factor hindering the development of telemedicine diagnosis. To address this problem, this paper proposes a secure decision tree classification scheme combining Chi-square (Chi-2), principal component analysis (PCA) and secret sharing (SS). It aims to provide secure telemedicine diagnosis services for cloud users while ensuring model and data privacy. The feature engineering of medical data through Chi-2 and PCA reduces computational and communication overheads; the storage and computational efficiency of the model is optimised using a pseudorandom generator (PRG); During the classification process, both the user and the cloud service provider operate the decision tree model in a secret-sharing manner, which guarantees the confidentiality of the model and the user's physiological data. Experimental results demonstrate that our scheme enhances accuracy by 2% over existing methods while safeguarding privacy, reduces time costs by 49.32%, and decreases storage overhead by 20.33%.