社会网络动态发布中防止标签邻居攻击的匿名方法研究

发布时间：2018-02-24 09:28

本文关键词： 社会网络动态发布隐私保护　出处：《广西师范大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,随着互联网技术的蓬勃发展,社交网络软件也越来越流行,如微博微信和Facebook等,这些社交网络应用为人们提供了方便的沟通平台,同时也产生了大量的有关用户的信息。这些信息具有广泛的用途,如进行广告投放、商品推荐和社会行为预测等。社会网络数据中包含大量的敏感信息,包括个人的属性信息(比如职业,薪酬等),个人的行为信息(比如个人的社交关系等),这些信息如果不进行处理就发布共享,有可能会侵犯用户的隐私。因此,社会网络数据发布的隐私保护问题成为众多研究者关注的热点。现有工作主要关注静态网络分析,但是很多应用涉及到网络的动态发展变化。与静态网络不同,面向动态网络数据的隐私保护提出了更高要求,它不仅是要保证某一时刻的数据满足匿名要求,还要保证多次发布隐私信息的安全,因为不同时刻之间的数据还存在内在的关联关系,攻击者可通过先后时刻发布的数据进行比对而获得更多的隐私信息。综上所述,现有的面向静态社会网络分析的隐私保护方法已不适用于动态发布的隐私保护。本文针对以上问题,提出了在社会网络动态发布中防止标签邻居攻击的隐私保护模型。主要的研究工作如下:首先,通过对社会网络动态发布中隐私保护方法的现状进行分析,指出了社会网络动态发布中的隐私泄露问题,并对现有用来解决社会网络动态发布中的隐私问题的隐私保护方法进行分析,发现其中存在的不足:1)现有的带敏感标签的个体的分组方法都是根据邻居标签相似度,没有考虑社会网络中的邻居结构信息,在对带敏感标签的个体的标签-邻居图进行模糊化处理的时候,使得社会网络图添加了很多噪声边,影响数据的可用性;2)虽然每次发布的社会网络都满足社会网络隐私保护的需求,但是攻击者仍然可以根据背景知识将多次发布的社会网络图进行关联分析,就可能唯一识别带敏感信息的个体;3)如果对整个图进行模糊化方法处理的话,不利于对数据进行分析,很多不需要保护的个体也进行了随机化处理,使得原有的社会网络结构信息损失增大,降低了数据的可用性。其次,针对社会网络动态发布的防止标签邻居攻击的隐私问题,本文提出了dynamic-l-diversity隐私保护模型。并对以下问题进行研究:1.针对现有分组方法中只考虑邻居标签信息,并未考虑原始社会网络中节点的邻居结构信息的问题,本文提出一种根据结构相似度对带敏感标签的个体进行分组的方法。2.针对通过比对动态发布中节点的标签邻居信息就能唯一识别带敏感标签个体的问题,本文提出在动态发布的社会网络中满足l-多样性的匿名方法。本文首先采用随机扰乱的方法来改变图结构,接着,随机添加删除边,然后采用仅对带敏感标签个体的邻居图进行随机化处理的方法,对社会网络图进行模糊化处理,使得每条边都有相应的概率存在于社会网络,同时使攻击者唯一识别带敏感标签的个体的概率不高于1/l0除此之外,本文仅对带敏感标签的个体进行保护,减少了不确定图的数量,提高了数据的可用性。最后,基于本文的动态网络中防止标签邻居攻击的隐私保护方法,提出了详细的、系统的设计方案。本文使用了三种评价指标,分别是平均度数、聚类系数、结构熵。评价结果表明,本文的社会网络动态发布中的隐私保护方法可以保留网络结构的属性特征,同时保证了个体的隐私信息的安全。
[Abstract]:In recent years, with the rapid development of Internet technology, social network software is becoming more and more popular, such as micro-blog, WeChat and Facebook, the social network application provides a convenient platform for people to communicate, but also produces a large amount of information about the user. It has a wide range of uses, such as advertising, product recommendation and social behavior prediction. Contains a lot of sensitive information in social network data, including the attribute of personal information (such as occupation, salary, etc.) the personal information behavior (such as personal social relations, etc.), this information if not handled on the release of sharing, may infringe on the privacy of users. Therefore, the problem of privacy the protection of social network data publishing becomes the focus of many researchers. The work mainly focus on the static analysis of the existing network, but many applications involve dynamic network. The static network, put forward higher requirements for dynamic network data privacy protection, it is not only to ensure that a moment of anonymous data to meet the requirements, but also to ensure the privacy of information security issued several times, because of the different between the data also have inherent relationship, an attacker can get more information than privacy the moment has released data. In summary, privacy protection method for static analysis of the existing social networks are not suitable for the dynamic release of privacy protection. In this paper, aiming at the above problems, put forward to prevent privacy protection model label release network dynamic neighbor attacks in society. The main research work is as follows: firstly, through the analysis of the status quo of social network privacy protection method in dynamic publishing, pointed out that the dynamic social network privacy issues in the publication, and is useful for The privacy protection method to solve the privacy problem in the dynamic social network analysis, find the problems: 1) packet method with sensitive label individual existing are based on neighbor tags similarity, without considering the neighborhood structure information in the social network, when fuzzy processing on with sensitive labels individual label - neighbor graph makes social network graph to add a lot of noise, affect the availability of data; 2) although each release of the social network to meet the needs of society to protect the network privacy, association analysis social network graph but the attacker can still according to the background knowledge will be released several times, can only identify the individual with sensitive information; 3) if the indistinct way on the map, is not conducive to the analysis of the data, many do not need to protect a body was also carried out with Machine processing, increasing the original structure of the social network information loss, reduce the availability of data. Secondly, according to the dynamic social network released to prevent privacy label neighbor attacks, this paper proposes a dynamic-l-diversity privacy protection model. And through the study on the following issues: 1. aiming at the grouping method only considering neighbor label information that does not take into account the neighbor node network structure information in the primitive society of the problem, this paper proposes a structure according to the similarity of individuals with sensitive label grouping method.2. for the unique identification label information can be released in dynamic neighbor nodes by comparing with sensitive labels for the individual, this anonymous method l- diversity to meet in the dynamic release of social network. Firstly, using the method of random disturbance to change the graph structure, then randomly add delete edge, Then the only method of randomization of sensitive label individual neighbor graph on the social network graph with fuzzy processing, so that each side has a corresponding probability exists in the social network, and allow an attacker to uniquely identify with sensitive labels for individual probability is not high in 1/l0 in addition, this article only carries on the protection with sensitive label individual, reduce the number of uncertain graph, improve the availability of data. Finally, the privacy protection method to prevent the network attack dynamic neighbor tags based on the proposed in detail, the design scheme of the system. This paper uses three indexes, respectively is the average degree, clustering coefficient the structure, entropy. The evaluation results show that the method of privacy protection issue in the dynamic social network can retain the character of network structure, but also ensure the individual privacy of information security.

【学位授予单位】：广西师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP309

【相似文献】