文章针对公开数据集上的隐私数据保护展开研究，分析了经典的k匿名算法在处理连续发布的数据集时存在的不足，在新的应用场景下对其进行改进。文章提出的算法通过增量式的数据处理技术减少了时间开销，适用于大规模数据集的快速连续发布。算法通过为每个数据元组选择最优等价类，有效控制了信息损失。算法以敏感属性值泛化技术代替了伪造元组的引入，保证了数据集上只包含真实数据，提高了数据集的可用性。通过实例分析发现提出的算法可以很好的解决连续发布数据集上的隐私保护问题。% In this paper, we did research for privacy protection in datasets published. We analyzed classic k-anonymity algorithms and their shortcoming in dealing with datasets published continuously. We improved the existed algorithm in new scene. Our new algorithm used incremental techniques to lower time cost. This made the algorithm good at dealing with serial large datasets. We chose best equivalence class for each tuple added and limited information loss. We used generalization of sensitive values to replaced introduction of counterfeit tuples. Datasets produced by new algorithm contained only real data. The new algorithm improved the usability of datasets published. After analysis of example, we can find that the improved algorithm can protect privacy in datasets published continuously.
展开▼
