Research On Attributes Clustering And Bidding Mechanism Based SaaS Oriented Privacy Preserving Method

Along with the development of cloud computing, SaaS is more and more popular. Softs are running at the side of provider as a service, so the users neither need to spend a lot of financial resources to buy the required software, nor to use a large amount of manual labor and energy to maintain the steady running of the software. This requires users upload their data even though it includes privacy information to the servers of provider, this accompanies problems of data security. How to solve the problem of privacy preserving effectively causes the research institutions, experts and scholars for concern.There exist two main kinds of privacy preserving methods, including encryption and confusion. Encryption uses the intractable problem or irreversible process mathematically as algorithms deforms the original data. Confusion uses the generalization or anonymous way to hide the true information. Generalization,expands the discrete data to a continuous interval. Anonymity use the fragmentation or sand mixing way to confuse data in each groups equivalent. However, encryption algorithm is designed complex to improve the privacy preserving degree, so the cost time is so long to be tolerated for SaaS application. Confusion is more efficient than encryption but it breaks the original information and could not rebuild, besides, it will produce dirty data.Addressing the challenges presented above, we propose a concept of data fragmentation. It divides tenants’ identity, such as name, identity card number or social security number, and privacy data, such as diseases, salaries into different data fragmentations, and confuses the corresponding relationship. It operates more efficient than encryption, at the same time, it can rebuild the original information based on the corresponding relationship, avoiding losing users’ data.The granularity of data fragmentation is finer, the data security degree is higher, however, the more time is cost for combination. How to divide the data properly that application can response as fast as possible when the users’ privacy is guaranteed. This paper statistics users’ access pattern and generates attribute association matrix, through BEA algorithm clustering the matrix, optimal privacy fragmentation strategy can be obtained at last. The application operation on the data fragmentation generated by this strategy needs the least combination times and performs better.Considering the actual situation, different users define different privacy, for example, healthy people define disease as no privacy and the public figures thought office number is not privacy. This paper proposed a concept of privacy constraints giving personalized, through two stage pricing mechanism, independent pricing algorithm and centralized pricing algorithm, controls users’ demand properly, we can maximize the utilization of resources while both provider and tenants are satisfied for the price.