Distributed Clustering Algorithm Based On Privacy Protection

As data mining technique has been becoming more and more important in commerce, a sensitive question which is privacy protect in data mining, is also attract more and more attention both of normal persons and researchers. PPDM (Privacy Protect Data Mining) is a hotspot of artificial intelligence now.In this article, we mainly study on the multi-security calculate problem. And the goal is to propose a clustering arithmetic, which can prevent every participator's subtle information, which contains both the sensitive data and the knowledge from data analyzing. In the privacy protect distributing clustering arithmetic, we adopt the concept of min-cluster from data string, and join in multi-security protocols. From the experiment, the PPDCA is proved to be with high accuracy, with low time complications, and the last but most important security.Considering the increment of information and cooperation, it is quite common that some other data owners may enter the cooperation system, which means all the results should be update. In order to avoid clustering all the data again, we bring forward the IDPPKMeans(Increment Distributing Privacy Protect K-Means). At last, the experimental results and analysis bear out the arithmetic has lots attributes which are needed in PPDM, such as high accuracy, low time and communication complications, and security with no data or private knowledge leakiness.