基于差分隐私的决策树发布技术研究

摘要

Privacy disclosure issue is becoming more and more serious due to big data.We proposed a differential private generalization data publishing algorithm for decision tree.The algorithm is based on the non-interactive model, in the process of attribute segmentation by combining similar branch to reduce the overall noise, and keeps more of original information, so it can improve the accuracy of classification.According to the need of exponential mechanism, adaptive allocation privacy budget, compared with the previous algorithms, under the condition of the same privacy budget it can make the data more differentiated, and decision tree classification accuracy is higher.The experimental results also prove the validity and superiority of this algorithm.%近年来,大数据所带来的隐私泄露问题日趋严重,如何在保护数据隐私的同时保留足够的信息进行数据分析是研究者面临的重要挑战.针对数据分析过程中可能产生的隐私泄露问题,提出一种基于差分隐私的决策树发布法,该算法是基于非交互模型,利用指数机制保证细分属性的选择满足差分隐私保护,根据数据集的特点自适应分配隐私预算,相比已有的算法隐私预算分配更合理,决策树的分类准确性更高.实验结果验证了本算法的优越性.