符合差分隐私的流数据统计直方图发布

A histogram publishing method for stream data satisfying differential privacy

直方图是一种被广为应用的统计数据发布形式,其潜在的隐私泄露风险是当前数据隐私保护领域的关注点.该文针对流数据的直方图发布问题,提出一种符合差分隐私保护要求的方法.其主要特点包括:(1)将w-事件引入流数据的直方图发布加噪机制以确保其满足差分隐私保护需求;(2)采用卡尔曼滤波方式对加噪后的流数据进行后置处理以改善数据效用;(3)通过指数平滑法改进卡尔曼滤波方式避免相邻数据之间的突变性.论文以UCI的两个真实数据集为基础进行流数据直方图模拟发布实验,结果表明该文方法在不同差分隐私预算约束、不同窗口大小情形下均具有明显优势,可在相同隐私保护水平下获得更高的数据可用性.

Histogram publishing is a widely-used statistical data publishing format,and its potential privacy leakage risk is the focus in the field of data privacy protection.In this paper,aiming at the issue of publishing histograms of stream data,we propose a method that meets the requirements of differential privacy protection.Its main features include:(1)Introducing w-event into the histogram distribution with noise enhancement mechanism of stream data to ensure that it meets the requirements of differential privacy protection;(2)Post-processing the noisy stream data by using Kalman filter to improve the data utility;(3)Improving the Kalman filtering method to avoid the abrupt change between adjacent data by the exponential smoothing method.Based on two data sets from UCI,the histogram simulation publishing experiments of streaming data are carried out.The results show that this method has obvious advantages under different differential privacy budget constraints and different window sizes,and can obtain higher data availability under the same privacy protection level.