基于深度学习算法与AP聚类的轻量级分布式数据泄露检测  

Lightweight distributed data leakage detection based on deep learning algorithm and AP clustering

作　　者：商圣光 SHANG Shengguang(Dongying Branch of Sinopec Group Shared Services Co.,Ltd.,Dongying,Shandong 257099,China)

机构地区：[1]中国石化集团共享服务有限公司东营分公司,山东东营257099

出　　处：《计算机应用文摘》2024年第18期79-81,共3页

摘　　要：轻量级分布式数据在分布上存在属性特征。在对该类数据的泄漏状态进行检测时,可能导致输出结果误差较大。因此,文章提出了基于深度学习算法与AP聚类的轻量级分布式数据泄露检测方法。结合数据簇不同区域meanshift模长的差异性,对分布式数据的局部中心量度进行了重新定义,并采用深度学习的方式将属于边缘区域的部分进行了剔除处理。同时,引入了AP聚类算法,将数据泄露检测问题转化为原始轻量级数据点初始连接能力计算和连接能力衰减状态计算的问题,从而根据聚类后数据簇内数据点之间的关系确定数据泄露状态。在测试结果中,设计检测方法的输出结果与实际设置值表现出较高的拟合度,对应的误差始终稳定在0.008以下,最小误差为0。Lightweight distributed data has attribute characteristics in distribution,and when the leakage state of this kind of data is detected,the output error will be large.Therefore,this paper proposes a lightweight distributed data breach detection method based on deep learning algorithm and AP clustering.Based on the difference of meanshift mode length in different regions of data cluster,the local center measure of distributed data is redefined,and the edge region is eliminated by deep learning method.At the same time,the AP clustering algorithm is introduced to transform the problem of data leak detection into the calculation of the initial connection ability and the decay state of the connection ability of the original lightweight data points,and the data leak state is determined according to the relationship between the data points in the cluster.In the test results,the output result of the designed detection method and the actual set value show a high degree of fit,the corresponding error is always stable below 0.008,and the minimum error is 0.

关 键 词：深度学习算法 AP聚类 轻量级分布式数据 泄露检测 MEANSHIFT 局部中心量度 连接能力 衰减状态 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]