A Differential Privacy Preserving Method for Adaptive k-nets Clustering

Abstract

The invention discloses a differential privacy protection method for adaptive K-Nets clustering. Firstly, the natural neighbors of all data points are obtained by calculating natural neighbors. When the total number of natural neighbors of all data points remains unchanged or the number of natural neighbors is 0 When the number is constant, the obtained K value is the parameter of the K nearest neighbors we need. Then use the K-Nets network model to calculate the KNN average distance of the data points as the score value of the data point. In order to protect privacy, the score value is added with Laplacian noise for protection. Then sort the scores and select the clusters with density from high to low, and then judge to find the naturally formed M clusters, and judge whether the points that have not been added to the M clusters are outliers, and there is no outlier if they are not Add the points in the cluster, and traverse these points into the cluster closest to it. The invention can effectively ensure that the privacy of data is not leaked.

Description

technical field [0001] The invention relates to the technical field of data privacy protection, in particular to an adaptive K-Nets clustering differential privacy protection method. Background technique [0002] The explosive growth, wide availability and huge amount of data make our era a real data age, so we urgently need powerful and general-purpose tools to find valuable data from these massive data and transform these data into Organized knowledge, this need led to the birth of data mining. In data mining, clustering is an important technology that has been continuously studied in recent years. Clustering is a process of dividing a set of data objects into multiple groups or clusters, so that objects within a cluster have high similarity but have low similarity with objects in other clusters. Clustering as a data mining tool has been rooted in many application fields, such as biology, security, business intelligence and web search. In 2018, loannis A. Maraziotis et ...

AI Technical Summary

Problems solved by technology

[0003] What the present invention is to solve is the problem that the K value parameter caused during the operation of the existing K-Nets clustering model has a great influence on the clustering results, and provides a differential privacy protection method for adaptive K-Nets clustering

Images