基于K-means方法的MapReduce模型社交网络隐私保护研究  

Social network privacy protection research based on the MapReduce model of K-means

作　　者：张娱嘉 张景璐[2] ZhangYujia;Zhang Jinglu(Zhiji Motor Technology Co.,Ltd.,Shanghai 201210,China;Beijing Polytechnic,Beijing 100176,China)

机构地区：[1]智己汽车科技有限公司,上海201210 [2]北京电子科技职业学院,北京100176

出　　处：《无线互联科技》2022年第20期162-165,共4页Wireless Internet Technology

摘　　要：传统隐私保护方法已逐渐无法应对多种背景下的恶意分析问题,文章研究了一种满足分布式环境下的差分隐私的算法。算法通过使用计算框架来控制主任务的迭代执行,每次分配子任务独立并行计算每个数据片中的每条记录与聚类中心的距离,标记其所属的聚类。分配子任务来计算同一簇中的记录数量和属性向量的总和,并利用机制产生的噪声干扰来实现隐私保护,从理论上证明了整个算法满足差分隐私保护。Aiming at the problem that traditional privacy protection methods cannot cope with malicious analysis under arbitrary background circumstance,an algorithm that satisfies differential privacy in a distributed environment is proposed.The algorithm uses a computing framework to control iterative execution by the main task;assign sub-tasks to independently and parallel calculate the distance between each record in each data piece and the cluster center and mark the cluster it belongs to;assign sub-tasks to calculate the records in the same cluster.The sum of the quantity and the attribute vector,and use the noise disturbance generated by the mechanism to realize privacy protection.According to the combined characteristics of differential privacy,it is theoretically proved that the whole algorithm satisfies differential privacy protection.Experimental results prove that this method guarantees better usability while improving privacy and timeliness.

关 键 词：K-MEANS MAPREDUCE 差分隐私 LAPLACE分布 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]