抽象的

A Secure Framework for Protection of Social Networks from Information Stealing Attacks

Dr. M.BalaGanesh Mrs.V.Sathya S.Vinoth Kumar

Social networks are online applications that allow their users to connect by means of various link types. Since these sites gather extensive personal information, there is a promising chance for leakage of personal information and inference attacks. To prevent the social networks from such attacks, a secure framework is proposed here. As first step, the social network is modeled as a connected graph where nodes and edges represent users of network and relationships among them respectively. Then three kinds of learning methods are applied for modeling the inference attacks. The sensitive attributes of each person‟s record is gathered by applying Naive Bayes Classification and each sensitive attribute is classified into a class set. This is known as local classification scheme. In relational classification scheme, relationship between nodes that is persons are examined and link information is inferred. Collective inference scheme attempts to use both the local and relational classifiers in a precise manner to attempt to increase the classification accuracy of nodes in the network. After modeling the network attacks, the sensitive attributes and friendship links are either removed or modified. Thus the social network is sanitized and removing details and links reduce the classification accuracy of classifiers. Thus the proposed approach effectively maintains confidentiality of the data set even after its release so that the attackers have no chance to infer sensitive information of users.