Neural Network based Privacy Preservation Data Mining for Social Network Sites: A novel framework for privacy-preserving data mining in social network sites using neural networks

Abstract

Due to increasing use of Online Social Networks (OSNs) applications such as Facebook, Twitter,etc., several research challenges related to security and privacy of OSN users introduced. The end users ofOSNs expecting that social networks should strong enough to preserve their private data secure from theattackers. In this research work, we introduce the novel data mining based framework to protect the OSNsfrom various privacy violation concerns. Risk of disclosure of individual’s confidential information haverisen tremendously due to widening of social network and publication of its data. From securityperspective privacy retaining becomes mandatory prior to service providers publish networkinformation. Recently, preservation of privacy in data of social networks has become most challengingand concerning problem as it has caught our lives in a dramatic way. Various methods ofanonymizationexist thathelps in retaining privacy of social networking.By developing graph and nodesdegree, k-Anonymity and among all available techniques is an utmost one that assist in deliveringsecurity of information on internet. With major manipulation in editing of node techniques in thisresearch paper, improvement of K-anonymity has been explained. With integration of same degree inone group, clusters are developed and processes are repeated untillrecognition and identification ofnoisy data is done. For minimizing node misplacement in groupsan Advanced Cuckoo Search iscommenced and processed. To cross verifystructure and for reducing node miss placement in groupsoutcome of Cuckoo Searches are combined with Feed Forward Back Propagation Neural Networks. Wehave computed the Information Loss and Average Path Length of proposed model. These resultsshowed that the reduction in these parameters to a good extent compared to other implementations.These values of Information Loss and Average Path Length in case of a network with 9 nodes areobtained as 0.24 and 32.2 respectively.