User Privacy Analysis And Protection For Social Network

With the fast development of network and information technology, Social Networking Services (SNS) are more and more popular. In recent years, a large number of users store the private information in the web based Social Network. Once these private data were leaked, the impact and damage would be more serious than the real world. Therefore, how to protect the privacy of social network has become a pressing problem.Access control is one of the important methods to protect privacy in Social Network. Based on various levels of user involvement, the current access control mechanism can be divided into default privacy、customizable privacy and adaptive privacy. The default privacy setting is the pre-configured privacy settings by SNS. The customizable privacy setting is based on default privacy setting which allows users to customize control strategy according to their needs. The adaptive privacy setting is automatically inferring the user’s privacy settings via analyzing extracted user input, feature information and context information. First two privacy settings require doing reasonable and accurate classification to visitors and private data. However, such method is not suitable for unclassified visitors and data. The method of adaptive privacy does not quantitatively analyze the relationship between user privacy-preference and extractive information. In the initial stage of inferring user privacy preferences, it may be more dependent on direct interaction with user.For problems mentioned above, we propose the analysis model for user privacy-preference based on probabilistic and the analysis model for accessing object based on user privacy-preference. The main contributions are as follows:To address the authorization of unclassified visitors, a probabilistic analysis approach is proposed to measure the relevance between existing users’privacy policies and visitors’ attributes. Based on this measurement, a new visitor can be classified into an appropriate group according to their attributes and get the associated access rights. For the problem of large number of object resource, we propose a combination of a probabilistic analysis approach and interactive learning method. Firstly the probabilistic analysis approach is proposed to measure the relevance between existing objects’labels and group authorization. On this basis, a new object can be authorized to an appropriate group according to their labels. However, there are too many objects to obtain a precise number and inaccurate labels. We need to fatherly analyze object permission assignment by the interactive learning. At the same time, from different angles we evaluate the above methods. Then experiments are performed to verify our methods.