Privacy Protection Based On Attributes Association Hiding

With the continuous development of network technology, electronic data processing becomes easy, so more and more data is collected, stored in the form of electronic records in the database or the cloud, thus, we enter the era of big data.The coming era of big data not only enrich and facilitate our daily life,but also provides data support to research of government and public service,at the same time take back lots of wealth. However, some data related to the user’s sensitive information inevitably. The development of data mining technology also increases the risk of user privacy information leakage. Users’ privacy disclosure has become one of the main obstacles to the rational use of big data.Faced with uncredible data storage providers, there is a series of problems and challenges to protect users privacy in publishing data:(1) the rapid development of social networks, resulting in the user information associated with each other,you can even get the sensitive information of one person from the publishing data of others perple,that means,you should take into consideration the social network when protecting privacy;(2) The user’s privacy may change with users recognization,the mechanism to protect the privacy of users should adapt to dynamic changes of user demand;(3)what is privacy is different from people to people, privacy protection process needs to reflect the individual needs of users;(4) privacy protection can not sacrifice the availability of data.(1) First, I in this paper reviews the existing privacy protection techniques and models,mostly focus on k- Anonymity and differential privacy protection.Then we conclude with the characteristics combining the current situation with the weak point of prior algorithm.(2) Due to the characteristic of muti-attribute of user information,wo introduce a attribute partition method. Some properties may not contain the user’s sensitive information, but by focusing on a combination of attributes can be inferred sensitive information will still leaking user privacy.By identifying the associated properties,storing the combining attribute separately to prevent privacy disclosure.(3) For the individual needs of users, we propose a user-defined privacy protection method.Users can define privacy by demand,and formulate a specific schema for privacy protection.(4) In order to ensure the availability of data, we introduce the classification model of data mining to define the availability of data after processed by the privacy protection technology,and compared with it before.