Privacy-preserving association rule mining for horizontally partitioned healthcare data: a case study on the heart diseases

Abstract

In recent years, a trend of electronic health record (EHR) system can be seen increasingly in the hospitals, which has generated huge amount of electronically stored data of patients. Association rule mining technique is very helpful in the numerous applications of healthcare (e.g., correlation between disease and symptoms, disease and offering effective treatment and predicting risks of disease based on the historical data, etc.). The data collected by an EHR system are very important for the medical research. Currently, a patient health report is derived on the basis of a physician's own experience and on the association rule mining results of a local EHR system maintained by a particular hospital. Association rule mining results will be more accurate if the data of all local EHR systems are integrated and association rule mining is performed. Integration of local EHR systems requires the sharing of local EHR data. Sharing of patient records violates the privacy of patients. Hence, medical research is focused on the problem of mining association rules without sharing of local private EHR data. Privacy-preserving distributed association rule mining (PPDARM) solves this issue by mining the association rules while preserving the privacy of patients. In this paper, an approach for the PPDARM is proposed for collaboratively performing association rule mining by all local EHR systems while preserving the privacy. The proposed approach is also analysed with the heart disease dataset.

Privacy-preserving association rule mining based on electronic medical system

“…Gkoulalas et al [ 18 ] presented a survey of algorithms for the patient data in EHR systems, which remains useful for subsequent analysis tasks in a privacy-preserving way. Privacy-preserving association rule mining for horizontally partitioned healthcare data [ 19 ] and vertically partitioned healthcare data [ 20 ] were both proposed by Domadiya and Rao respectively. Yigzaw et al [ 21 ] proposed a feasible architecture which can protect the privacy of the patients, clinicians, and healthcare institutions as well as mine the clinical performance of a clinician over the patient data from EHR systems.…”

“…Domadiya et al [ 19 ] proposed a PPDARM scheme for horizontally partitioned healthcare data with insecure communication channel. They also claimed that the scheme can produce an accurate mining result without compromising the data privacy of every EHR system.…”

“…However, we find the scheme has serious security flaws in protecting the data privacy of every EHR system. The improper points in [ 19 ] are summarized as follows. Every EHR system (participant) is required to make interactions with the two neighboring EHR systems.…”

“…Moreover, the higher communication overhead is also an issue to be improved. The performance analyses of the PPDARM scheme in [ 19 ] are not exactly correct. Take the set-up phase for example.…”

Secure Association Rule Mining on Vertically Partitioned Data Using Private-Set Intersection

“…Multiple schemes [3][4][5][6][7], [27], [28] have been proposed for a single data owner to outsource ARM. In a distributed data environment involving multiple data owners, there are schemes for outsourcing in a horizontal [13], [15], [29][30][31], and vertical distribution environments [32][33][34][35].…”