一种改进的差分隐私聚类位置保护方法

An Improved Differential Privacy Clustering Location Protection Method

差分隐私常被应用于位置隐私保护场景中,通过给位置点加入干扰噪声来混淆真实数据以达到保护隐私目的,但该方法会造成大量噪声数据冗余,影响位置的真实性。为解决该问题,提出一种新的基于差分隐私的DPK-MO算法来保护用户真实位置。在确定初始中心点时加入邻接密度和最小误差平方,并始终选取样本误差平方和最小的点作为中心再聚类,剔除离散点,合并密度小的聚类集,最后合理加入符合差分隐私的拉普拉斯噪声来得到虚拟位置。实验结果证明,该方法可有效缓解数据集范围广、边界值影响大、密度分布不均的问题,降低了查询误差。在同一隐私参数下与差分隐私K-means聚类方法相比,数据可用性提升了30%。

Differential privacy is often used in location privacy protection scenarios.It mainly confuses the real data by adding interfer⁃ence noise to the location points to achieve the purpose of privacy protection,but this method will cause a lot of noise data redundancy and affect the authenticity of data location.In order to solve this problem,a new dpk-mo algorithm based on differential privacy is pro⁃posed to protect the user's real location.When determining the initial center point,the adjacency density and the least square error are added,and the point with the least square error of the sample is always selected as the center for re clustering.The scattered points are removed,and the clustering sets with small density are merged.Finally,the Laplacian noise which conforms to differential privacy is reasonably added Sound to get the virtual location.Experimental results show that this method can effectively alleviate the problems of wide range of data sets,large influence of boundary value and uneven density distribution,and reduce the query error.Under the same privacy parameter,compared with the differential privacy k-means clustering method,the data availability is improved by 30%.