Research On Privacy-Preserving Data Mining

With the development and the further application in human life, data mining is affecting our life day by day. But it also offends our privacy while it brings the knowledge and makes convenience to us. At the same time, people pay more attention to their own privacy, and it makes more difficult to complete data mining task. Privacy-Preserving Data Mining has emerged to solve this problem. It aims at bridging the gap between privacy-preserving and knowledge discovery.First of all, the paper summarizes the typical algorithms of resent research on privacy-preserving data mining, and then it points out the problem that the privacy -preserving classification algorithms of centralized data only suit on binary attribute or the data must satisfy some certain probability distribution and it does not suit on categorical attribute data; Second, the paper introduce a new algorithm that suits on categorical attribute, this algorithm uses the method of generating random data to protect original data and it modifies the way of calculating information gain ratio to accomplish classification mining. This algorithm can deal with categorical data and does not require the data follow some certain probability distribution. At last, the experiment shows the algorithm is correct.