Background/Objectives: To develop a kNN-privacy preserving model for preserving the privacy of the patients in a cloud assisted e-healthcare system as the sensitive information is needed to be maintained confidential and should not be revealed to public users other than the physicians.
Methods/Statistical analysis: PPDM uses a privacy-preserving fully Homomorphic data aggregation as the basic scheme. The outsourcing of disease modeling and the early intervention is performed by developing privacypreserving function correlation matching PPDM1 from dynamic medical text mining and also a privacy-preserving medical image feature extraction PPDM2. Both PPDM1 and PPDM2 provides higher security level with reduced cipher text attach possibility and minimal overhead. Though the computational and communication overhead are reduced in PPDM, the use of correlation function threshold in PPDM1 can further be improved by utilizing an efficient machine learning algorithm. Hence, the simplest and efficient machine learning algorithm, k-nearest neighbor is utilized to develop kNN-PPP model.
Findings: In kNN-PPP model, instead of using correlation function threshold based matching, secure squared Euclidean distance of encrypted personal data and encrypted physician template is determined and then matched with better probability. Secure squared Euclidean distance protocol and secure multiplication protocols are the most prominent protocols among those utilized in kNN-PPP model.
Improvements/Applications: Using kNN-PPP protocols, the computation and communication overheads are also reduced considerably than the PPDM model for the better health status determination of the patients. Experimental results also show that the kNN-PPP model has minimized overheads and higher matching probability.