Research Of Outlier Detection Base On Privacy Preserving

With personal privacy have become more concerned in modern society,privacy protected data mining has become an important research direction. Now, privacy protected data mining as a new research direction, it have has broad application prospects in shared many government documents, medical institutions, joint research in areas such as e-commerce. Privacy protected data mining is a set of contradictions. While it should mine a lot of useful information and data models from huge data, on the other hand, it is necessary to keep those privacy data from each party to participate secretly.There are two methods used in privacy protected data mining. One is to data disturbance technology, and another is to use secure multi-party computation. The former mainly focuses on centralized data mining, another focuses on distributed data mining. According the concept of privacy in data mining, goal of privacy protected data mining is given. With introducing data distribution method,privacy protection technology,data mining task and commonly data mining algorithms, current existing privacy protected data have a more comprehensive review.The general principles and typical data mining technology of data mining firstly, especially focused on secure multi-party computation technology. Then, base on oblivious transfer protocol and homomorphic public key cryptosystems secure multi-party computation is mainly analyses. There is conspiracy of multi party in those protocols. In order to deal with conspiracy problem, a more improved Secure Multi-party Addition protocol in terms of equal access to information for all parties is proposed.Outlier detection is used to find out significantly different outlier from other in dataset. It is a very important data mining technology. However, in current outlier detection base on privacy protected, if there is a conspiracy between some parties, privacy data of other parties will be leaked. So, with new secure multi-party addition protocol, a new outlier diction algorithm is given, it can get outlier based on distance from in the vertical division of distributed database. And there is not information divulged by conspiracy.