| Recently, with the development of Internet and various information exchange platform, people from the real world has formed a variety of social networks in different virtual spaces. Some exchange forums have collected a huge amount of users’ information and the relationships among users can also be obtained from the social platforms such as the Micro-blog or QQ. Social networks not only provide a lot of convenience, but also bring a lot of security risks to people’s lives. The disclosure of data privacy brings a lot of unexpected troubles to people’s lives. On the one hand, we want to enjoy the convenience of social networks. On the other hand, we don’t like the full transparency of our personal data. Therefore, the research of social network privacy protection method which is based on the differential privacy has great social significance and commercial value.The research content of this thesis is to solve the problem that traditional differential privacy protection method can only be applied to non-interactive social networks, but does not work well in interactive social networks. The main work of this thesis includes: first, due to the weakness of traditional differential privacy protection method, the thesis designs a differential privacy protection mechanism for interactive social networks and proves that it satisfies the requirements of differential privacy. Second, by considering the nature of social networks and the complexity of privacy attacks, we propose three practical indicators(query function) as the primary goal to protect privacy of social networks. In fact, the degree distribution, the cut set and the shortest path are the most basic components of social networks, and they should be protected against attackers. Third, according to the relevance of information, the thesis proposes the method of adding Laplace noise based on interactive factor. The corresponding interactive factor(parameter t) is calculated according to the degree of each node and then t is used to obtain noise in accordance with the Laplace distribution. Through the introduction of interactive factor, the interactive mechanism not only limits the effect of the node itself to global data, but also considers the relationship between the node and other nodes. Therefore, it can play an overall protection to release data and improve the corresponding protective effect.Through the experiments testing some preserving mechanisms, this thesis analysis the performance which is evaluated by the most basic nature of social networks and verifies the validity of the interactive differential privacy mechanism by comparing some factors such as the running time. The experimental results show that the interactive differential privacy mechanism proposed by this thesis can not only protect private data in the maximum extent, and at the same time, achieve a good protective result because of new interactive factors which give a stronger blinding to the protection of global data. |
