Data Mining Technology for Privacy Protection in Distributed Scenarios

Abstract

In the Internet era, data mining is an important means to seize users. However, data exists on different platforms, which are incompatible with each other, and user privacy is easily leaked when mining data. To address this issue, a distributed data mining method based on differential privacy is proposed. The method aggregates frequent itemset data from the top m items of branch nodes through a central node. The decision tree algorithm is used as a data classification method to set privacy budgets, optimize count queries, and perform importance attribute filtering. The experimental results showed that the improved algorithm had an average increase of 0.1 in data mining accuracy, an average increase of 0.115 in relative error, and an average decrease of 0.08% in privacy leakage probability. The data classification accuracy of the improved algorithm increased by an average of 0.28, and the privacy leakage probability during data classification decreased by an average of 0.06%. From this, the improved algorithm can significantly improve the accuracy of data mining and classification, significantly reduce the privacy budget required for data mining and classification, reduce the probability of privacy leakage, and greatly improve the security of user data.