基于随机块模型的社区隐藏统一框架  

A Unified Framework for Community Hiding Based on Stochastic Block Model

作　　者：刘栋 刘侠[1] 贾若雪 张文生 Liu Dong;Liu Xia;Jia Ruoxue;Zhang Wensheng(College of Computer and Information Engineering,Henan Normal University,Xinxiang,Henan 453007;Henan Provincial Key Laboratory of Artificial Intelligence and Personalized Learning in Education(Henan Normal University),Xinxiang,Henan 453007;Big Data for Teaching Resources and Education Quality Evaluation Henan Engineering Laboratory(Henan Normal University),Xinxiang,Henan 453007;Institute of Automation,Chinese Academy of Sciences,Beijing 100190)

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007 [2]教育人工智能与个性化学习河南省重点实验室(河南师范大学),河南新乡453007 [3]教学资源与教育质量评估大数据河南省工程实验室(河南师范大学),河南新乡453007 [4]中国科学院自动化研究所,北京100190

出　　处：《计算机研究与发展》2024年第7期1850-1862,共13页Journal of Computer Research and Development

基　　金：国家自然科学基金项目(62072160)。

摘　　要：社区检测是复杂网络分析的重要工具之一,可帮助深入了解网络的社区结构和节点间潜在的关系,但同时也带来了隐私泄露问题.社区隐藏作为社区检测的伴生问题,旨在以最小的边扰动代价破坏网络的社区结构,近年来受到越来越多学者的关注.但现有的社区隐藏方法忽略了网络的生成机制且缺少针对不同尺度隐藏的统一框架,因此提出了一种基于随机块模型的社区隐藏(community hiding-stochastic block model,HC-SBM)算法,该算法从网络生成机制角度构建了社区隐藏的统一框架,即实现微观(个体)、介观(社区)、宏观(网络)3个尺度上的社区检测算法攻击.其基本思想是基于随机块模型刻画网络的生成机制,特别是网络社区形成和分裂的规律和模式,挖掘生成过程中的关键性链接以及链接集合,最终通过最小代价扰动策略破坏网络社区结构.通过在真实网络上的大量实验,并与4种先进的基准算法进行比较,表明了提出的HC-SBM算法在社区隐藏效果更优.As one of the important tools for complex network analysis,community detection can be used to help gain insight into the community structure of the network and the potential relationship between nodes.However,it also brings privacy leakage problems.As a concomitant problem of community detection,community hiding aims to destroy the community structure of the network with minimal edge disturbance cost,and it has received more and more attention from scholars in recent years.However,the existing community hiding methods ignore the network generation mechanism and lack a unified framework for hiding at different scales.Therefore,we propose a community hiding algorithm based on a stochastic block model(HC-SBM),which constructs a unified community hidden framework from the perspective of network generation mechanism,and launches three-scale attacks against community detection algorithm,namely,micro(individual),mesoscopic(community),and macro(network).The principle of this method is to illustrate the generation mechanism of the network based on the stochastic block model,especially the rules and patterns of the formation and division of the network community,mining critical links and link collections in the process of network generation.Finally,the network community structure is destroyed at the minimum cost of perturbation.Through extensive experiments on real networks and comparisons with several mainstream baseline algorithms,the proposed HC-SBM algorithm is shown to be superior in terms of community hiding effect.

关 键 词：社区隐藏 社区检测 随机块模型 生成机制 社会网络分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]