Multiview Point Based Cuckoo Search Clustering Algorithm for Privacy Preserving on Multiple Sensitive Attributes with Horizontally Partitioned Data

Developing data mining techniques both suitable for databases as well as to maintain the individual privacy becomes the primary aim of research based on privacy preserving data mining (PPDM). The present PDDM clustering methods perform privacy preservation only with single point of view in which every tuple in the data matrix for data holder containing single sensitive attribute is been considered. In case of the clustering approaches based on multi-view point clustering, multiple sensitive attributes in a tuple has to be considered which remains inattentive. Considering these drawbacks, present study aims on multiple sensitive attributes (MSA) for observing new privacy risks and to investigate multi-view point based clustering methods in case of unknown data. Prior to solving that data disambiguation problem using Ramon-Gartner subtree graph kernel (RGSGK), the weight values are assigned for determining the kernel value for disambiguated data. The privacy is then gained from RGSGK for converted data matrix samples followed by creation of secure key for each data holder matrix with the help of Improved Ron Rivest, Adi Shamir and Leonard Adleman (IRSA). The proposed framework based on the multiview point based cuckoo search algorithm clustering is a novel context for tackling the problem of privacy preservation in the multiple sensitive attribute (MSA). This work constitutes distance, similarity and dissimilarity matrix for carrying out the clustering process. Comparison of the experimental result of proposed MVCSA Clustering algorithm with conventional methods has been done in terms of the F Measure, running time, less privacy and utility loss, communication cost for UCI machine learning datasets such as adult dataset and house dataset.