Research On View Publishing Method For Resisting Multi-Connection Attack

With the rapid development of infomation technology, data exchange frequency and quantity increase continually. As an effective method for information exchange, view publishing provide convenience for data exchange and data sharing, but the problem of sensitive information disclosure in the view publishing process are more serious, and becomes a new research subject in database security. This paper analyzes the current situation of the domestic and international security problem in view publishing process, and researches for the problem of security deciding of views.At first, several types of sensitive information are introduced via general classification and the number of connections. Several different attack methods, such as multi-party collusion and information inference, are introduced in detail. And each attack method is explained through an example. Based on this, the current methods of security deciding of views are analyzed, including conditional probability method, k-anonymity method and query-answering law, which lays a theoretical and practical foundation for follow-up study.Secondly, on the basis of the above research and existent solutions, a method for protecting multi-sensitive attributes based on maximal-leaf-subtree-first strategy is proposed. The multi-sensitive attributes privacy protecting problem is transformed into multi-sensitive attributes l-diversity problem. This method can effectively assure the security of the publishing views, otherwise, since the information loss is taken into account, this method not only protects privacy from leaking but also improves data accuracy.Moreover, the issue of security deciding of views is studied when they do not contain quasi-identifier or contain only a part of quasi-identifier attributes. When the views to be released include identifier attributes, the probability of view is estimated according to the principle of maximum-entropy. From the absolute security and relative safety point of view, the way of security deciding in publishing views is put forwarded based on the conditional probability algorithm. When the views to be released do not contain identifier attributes, the algorithm of security deciding in publishing views is proposed based on quasi-identifier. When there is information leakage, the information leak measurement methods and information disclosure elimination algorithms of the the above two cases are given.Finally, the multi-sensitive attributes protecting method and the view security deciding methods are analyzed by instances and experiments. The feasibility and validity of algorithms are proved.