Privacy Preserving Data Mining for Social Networks Ms. Brinal Colaco M.E. Student in Computer Engineering Dept. Faculty of Computer Engineering St. Francis Institute of Technology Mumbai University Mumbai-400103. India. Mr. Shamsuddin Khan Assistant Professor, Department of Computer Engineering St.Francis Institute of Technology, Mumbai University Mumbai- 400103, India Abstract—Advances in technology has made it possible for hackers and intruders to collect personal and professional data about individuals and the connections between them, such as their email correspondence and friendships on the internet. These hackers could either be third party agencies or individuals who are interested in knowing more about the users of the social networks. Most of the information present in these social networks is not private, yet learning algorithms could be used on the public data to predict private information. This paper focuses on the problem of private information leakage from the information present on the social networks. It represents the cause-effect relationships within social network data by the application of the soft computing technique of fuzzy Inference Systems. Sanitization techniques that could be used in various inference attack scenarios are suggested and effectiveness of these sanitization techniques is analyzed. Keywords—Social Network Security, Inference attacks, Fuzzy Inference System I. INTRODUCTION Social network analysis is the methodical analysis of social networks. In social networks, the individual entities within the networks are considered as nodes and the relationship between the entities are considered as links between the nodes [1][2].Various details like the hobbies, music interests, movies, books, favorite activities, professional information, etc. of the user is present in the user profile of the individual[3]. Social networks gather extensive personal information because of which the application providers have the opportunity of using the present information. However, in practice, privacy concerns can prevent these efforts [4][5]. The conflict between the desired use of data and individual privacy presents an opportunity for privacy-preserving social network data mining i.e. the discovery of information and relationships from social network data without violating privacy of the individual [3].In this research, the focus is on predicting private information using public information of the user that is present on the social network. In a social network, users have profile data that make certain aspects of their personality predictable. The identification or prediction of the underlying private attributes could have negative repercussions. For example, it is possible to determine a user‟s sexual orientation by obtaining few details from Facebook like user‟s gender, the gender they are interested in and the same details of the friends in their Friend list, or a user showing interest in a few groups on the social networks can predict his/her affiliation to a certain political group [6] [7].The inference of private information from the information present on the social networks is a major security breach. For this, a particular class of attacks are explored, namely, the class of Inference Attacks. These attacks are used to gain knowledge about a subject or a database. The attacks make an attempt to deduce sensitive information from trivial information that is publically available. The main goal of this work is to present a method based on the Fuzzy Inference System (FIS) that can be applied for the development of an expert system for predicting private information of an individual and also suggest ways to avoid the inference attack. No previous work has been done on implementation of FIS techniques to handle the specific problem for inference attacks on social networks. Efficiency of this technique will be compared to the efficiency of other techniques of Inference Attacks on Social Networks (e.g. Naïve Bayesian Classification technique) II. RELATED WORK The area of privacy within a social network is very vast and covers many research areas. Hay et al. [8] considers several ways of anonymizing social networks but the research mentioned in this paper is related to inferring the details on the social network not individually identifying individuals.Other authorshave tried to infer private information of the social networks. In [8], He et al, consider ways to infer private information via friendship links by creating a Bayesian network from the links inside a social network. Many inference attacks can be performed using third party extensions. In [3], Heatherly et al have workedon International Journal of Engineering Research & Technology (IJERT) Vol. 3 Issue 8, August - 2014 ISSN: 2278-0181 www.ijert.org IJERTV3IS080917 (This work is licensed under a Creative Commons Attribution 4.0 International License.) 1193