针对数据隐私保护直方图发布中数据划分不准确,以及所发布数据精确度较低的问题,提出了差值保序直方图数据发布方案.首先,通过直方图重构算法找出直方图最优分组方案,将数据分组,并用平均值代替组内元素值;然后,对分组后的数据添加噪声干扰,使其满足差分隐私保护要求;再后,运用差值保序算法降低误差,并还原数据原始排序特征;最后,通过实验证明所提方法与保序直方图发布方法相比,可以在保证用户隐私不被泄露的情况下,准确划分数据、提高数据精确度及所发布数据的实用性.%Aiming to solve the issue of inaccurate data partitioning in data privacy protection, a difference isotonic regression histogram algorithm is proposed.Through the histogram reconstruction, the optimal grouping scheme of the data could be found, then, the group element values should be replaced by average value;then, adding noise to the data packet to satisfy differential privacy protection requirements;finally, D-value isotonic regression algorithm was used to reduce the error and reduction the original ranking feature.Experimental results show that the method can be used to divide the data accurately, and improve data accuracy and the usefulness of the data released when the user's privacy is not leaked.
展开▼