Preserve data-while-sharing: An Efficient Technique for Privacy Preserving in OSNs

Abstract

Online Social Networks (OSNs) have become one of the major platforms for social interactions, such as building up relationships, sharing personal experiences, and providing other services. Rapid growth in Social Network has attracted various groups like the scientific community and business enterprise to use these huge social network data to serve their various purposes. The process of disseminating extensive datasets from online social networks for the purpose of conducting diverse trend analyses gives rise to apprehensions regarding privacy, owing to the disclosure of personal information disclosed on these platforms. Privacy control features have been implemented in widely used online social networks (OSNs) to empower users in regulating access to their personal information. Even if Online Social Network owners allow their users to set customizable privacy, attackers can still find out users’ private information by finding the relationships between public and private information with some background knowledge and this is termed as inference attack. In order to defend against these inference attacks this research work could completely anonymize the user identity.

This research work designs an optimization algorithm that aims to strike a balance between self-disclosure utility and their privacy. This research work proposes two privacy preserving algorithms to defend against an inference attack. The research work design an Privacy-Preserving Algorithm (PPA) algorithm which helps to achieve high utility by allowing users to share their data with utmost privacy. Another algorithm-Multi-dimensional Knapsack based Relation Disclosure Algorithm (mdKP-RDA) that deals with social relation disclosure problems with low computational complexity. The proposed work is evaluated to test the effectiveness on datasets taken from actual social networks. According on the experimental results, the proposed methods outperform the current methods.