Privacy Preserving Distributed Association Rule Research Articles

Privacy Preserving Distributed Association Rule Mining Approach on Vertically Partitioned Healthcare Data

The trends of data mining in the healthcare is increased due to the digitization of healthcare with electronic health record (EHR) systems. This generates a huge amount of data on daily basis. Data mining with the healthcare data has given the new direction to medical research for early detection of diseases and improving patient care. Many data mining applications require the integration of data from the different sources. For example, the integration of outpatient medical records and health examination data helps to identify the correlation between abnormal test result and disease. The result of association rule mining on this integrated data helps to build the knowledge center for disease prevention, which facilitate the healthcare provider in follow up treatment and prevention. The integration of data requires the sharing of sensitive information about the patients. Disclosing the sensitive information violates the privacy of patients. In this paper, we tackle the problem of privacy preserving association rule mining in vertically partition healthcare data. Furthermore, we analyze the proposed approach in terms of privacy preservation, communication and computation cost.

Privacy Preserving Distributed Association Rule Hiding Using Concept Hierarchy

Data Mining enables important knowledge to be extracted from the data. This has made data mining an important and useful emerging trend It may further be possible that data is distributed among various parties. This raises the issue of privacy. There are many data mining techniques. Association rule mining is one data mining technique and is receiving much attention from the researchers so as to enable them to find existing correlations between data items. In a distributed environment it may be important to find global association rules. The global rules has to be calculated securely without disclosing the individual information of one party to other parties. This paper uses secure association rule mining over horizontally and vertically distributed data. Further, the global rules calculated may be sensitive to some individual parties and there arises the need for association rule hiding. A method of concept hierarchy is used to hide the sensitive association rules.

Analysis and Evaluation of Novel Privacy Preserving Techniques for Collaborative Temporal Association Rule Mining Using Secret Sharing

With digital data being abundant in today's world, competing organizations desire to gain insights about the market, without putting the privacy of their confidential data at risk. This paper provides a new dimension to the problem of Privacy Preserving Distributed Association Rule Mining (PPDARM) by extending it to a distributed temporal setup. It proposes extensions of public key based and non-public key based additively homomorphic techniques, based on efficient private matching and Shamir's secret sharing, to privately decipher these global cycles in cyclic association rules. Along with the theoretical analysis, it presents experimental results to substantiate it. This paper observes that the Secret Sharing scheme is more efficient than the one based on Paillier homomorphic encryption. However, it observes a considerable increase in the overhead associated with the Shamir's secret sharing scheme, as a result of the increase in the number of parties. To reduce this overhead, it extends the secret sharing scheme without mediators to a novel model with a Fully Trusted and a Semi Trusted Third Party. The experimental results establish this functioning for global cycle detections in a temporal setup as a case study. The novel constructions proposed can also be applied to other scenarios that want to undertake Secure Multiparty Computation (SMC) for PPDARM.