K-Q:支持海量查询的隐私泄露检测算法

K-Q:algorithm for privacy disclosure detection supporting large number of queries

针对推理检测需要的所有历史查询结果的数据规模较大这一问题,K-Q算法结合K-匿名模型在历史查询结果的存储规模上进行了优化,通过推理攻击模拟算法在线检测恶意查询。在真实数据集上的实验证明了K-Q算法可以自适应于查询规模的增长,在准确率和内存消耗上都明显优于已有的直接基于相关元组合并优化的T-D算法。

A key problem remained that the data set required to detect inference attack cannot all fit in memory, K-Q algo- rithm optimized the real data storage for each history query based on K-anonymization model, it detected the illegal query online through simulating the real inference attack. Experiments on real data demonstrate that K-Q algorithm can scale on query size, and perform on detect accuracy and memory consumption is better than the existed T-D algorithm which directly merge related tuples and also assure the privacy control＇ s granularity.