面向社会网络的隐私保护关键技术研究

发布时间：2018-12-19 08:54

【摘要】：随着网络技术以及社交网站的迅速发展,例如Facebook、Twitter、人人网等,通过社交网站进行交友、联系和互动的用户群体数量迅速增加。为了挖掘社会网络中的科研和商业价值,越来越多的研究学者和开发人员将其科学研究和应用开发的注意力集中到社会网络这种虚拟世界当中,社会网络分析已经成为社会学、地理学、经济学、信息学等诸多学科的研究热点。基于社会网络数据进行数据挖掘和分析潜在模式比传统关系数据更加科学、效果更好,然而,社会网络数据中包含敏感隐私信息,因此在数据发布和共享过程中需要对社会网络中的隐私信息进行保护。在社会网络中,隐私信息类型更加广泛,隐私泄露方式呈现多样性,使得防止社会网络中的隐私泄露具有很大挑战。保护社会网络隐私是数据隐私保护领域亟待解决的热点问题,需要针对不同的隐私信息类型和泄露方式来研究相应的保护技术。具体的,本文对多种社会网络隐私保护关键技术进行了深入研究,包括保护结点身份、敏感关系、敏感属性值等多种隐私信息,保持匿名图的数据可用性,本文贡献主要包括以下几个方面：(1)在结点隐私保护方面,研究了攻击者将加权社会网络图中的边权重作为背景知识来发动结点识别攻击、从而导致结点隐私泄露的问题。提出加权图结点隐私保护模型来防止基于边权重的结点识别攻击,并设计了泛化匿名方法(记作GA)来实现加权图结点隐私保护模型。实验结果证明,提出的加权图结点隐私保护模型可以有效地防范针对加权图的结点识别攻击,同时基于匿名图进行采样可以无偏地恢复原图结构性质。(2)在敏感关系隐私保护方面,攻击者可以采用链接推演技术来识别敏感关系,研究了如何阻止链接推演攻击导致的敏感关系隐私泄露。定义了两种链接推演攻击,单步链接推演攻击和级联链接推演攻击。为了阻止链接推演攻击,提出了一种基于链接世系溯源的防推演机制来切断敏感链接的推演路径,并设计了防推演算法,在保护敏感关系的同时保持了图数据可用性。实验结果证明,敏感链接防推演机制可以有效地保护社会网络中的敏感关系隐私,并保持了发布图数据的高可用性。(3)在敏感属性值隐私保护方面,考虑了复杂社会网络中的每个结点包含与之相关的个人信息的情况,研究如何防御社会网络的个人信息隐私泄露。针对此问题,设计了k-混淆(k-obfuscation)模型来保护个人信息隐私,并提出一种安全结点-个人信息映射机制,记作k-映射(k-mapping)。同时,设计了优化技术来提高k-映射的执行效率和数据可用性。通过实验证明,提出的k-映射方法在保护个人信息隐私的同时,降低了匿名过程所导致的个人信息损失、相近信息损失,使得匿名图数据具有高查询准确度。(4)在保持图数据可用性方面,研究了如何在图匿名过程中保持结点间的可达性。提出了可达性保持图匿名化算法(简称RPA算法),RPA算法的基本思想是将结点进行分组并采取贪心策略进行匿名,从而减少匿名过程中的可达性信息损失。为了提高RPA算法的执行效率,首先提出采用可达区间来高效地评估边添加操作所导致的匿名损失；其次,通过构建候选邻居索引,加速了RPA对每个结点的匿名过程。通过大量的实验分析,表明RPA算法生成的匿名图保持了结点间可达性,使得匿名图在可达性查询方面具有很好的数据可用性。(5)实现社会网络数据安全发布原型演示系统SNSPDEMO。SNSPDEMO系统可以针对不同隐私泄露类型对社会网络进行安全性检测,通过图形化接口直观显示存在隐私泄露的结点和边的信息；SNSPDEMO系统整合了本文中的社会网络隐私保护技术,从而生成提供相应隐私保护的安全社会网络图,并通过图形化接口显示系统所做的图修改操作,对比原图与安全图之间的差别。总之,本文从社会网络隐私保护中的潜在威胁和挑战出发,针对社会网络隐私保护的关键技术展开研究,如结点隐私保护、敏感关系隐私保护、敏感属性值隐私保护、保持安全图可用性等,从而为社会网络隐私信息提供更加全面和完善的保护奠定了基础。
[Abstract]:With the rapid development of network technology and social networking sites, such as Facebook, Twitter, and Renren, the number of user groups that make friends, contacts, and interact through social networking sites has increased rapidly. In order to tap the scientific and commercial value of social network, more and more researchers and developers focus their attention on the development of scientific research and application to the virtual world of social network, and social network analysis has become the sociology, geography, economics, The research focus of many subjects such as informatics and so on. Data mining and analysis of potential patterns based on social network data are more scientific and more effective than traditional relational data. However, the social network data contains sensitive privacy information, so privacy information in the social network needs to be protected during data distribution and sharing. In the social network, the privacy information type is more extensive, and the privacy leakage mode presents the diversity, so that the privacy leakage in the social network is prevented from being a great challenge. The protection of social network privacy is a hot issue to be solved in the field of data privacy protection. In this paper, the key technologies of the privacy protection of various social networks are deeply studied, including the protection of many kinds of privacy information such as the identity of the node, the sensitive relation and the sensitive attribute value, and the data availability of the anonymous graph is kept. The contribution of this paper mainly includes the following aspects: (1) In the aspect of node privacy protection, this paper studies the problem that an attacker can launch a node identification attack by using the edge weight in the weighted social network diagram as the background knowledge, thus leading to the problem of node privacy leakage. In this paper, a weighted graph node privacy protection model is proposed to prevent the node identification attack based on the edge weight, and a generalized anonymous method is designed to implement the weighted graph node privacy protection model. The experimental results show that the proposed weighted graph node privacy protection model can effectively prevent the node identification attack for the weighted graph, and the original structure property can be restored unbiasedly based on the anonymous graph. (2) In the aspect of the privacy protection of the sensitive relation, the attacker can use the link deduction technique to identify the sensitive relation, and study how to prevent the sensitive relation privacy leakage caused by the link deduction attack. Two link deduction attacks, single-step link deduction attacks, and cascade link deduction attacks are defined. In order to prevent the link deduction attack, an anti-deduction mechanism based on the tracing of the link world is proposed to cut off the deduction path of the sensitive link, and the anti-deduction algorithm is designed, and the data availability of the graph is maintained while the sensitive relation is protected. The experimental results show that the sensitive link anti-deduction mechanism can effectively protect the privacy of the sensitive relation in the social network and maintain the high availability of the published graph data. (3) In the aspect of the privacy protection of sensitive attribute values, consider how each node in the complex social network contains the personal information related to it, and study how to defend the personal information privacy leakage of the social network. In this paper, the k-aliasing model is designed to protect the privacy of personal information, and a security node-personal information mapping mechanism is proposed, and k-mapping is recorded. At the same time, the optimization technique is designed to improve the implementation efficiency and data availability of k-map. The experiment shows that the proposed k-mapping method reduces the personal information loss and similar information loss caused by the anonymous process while protecting the personal information privacy, so that the anonymous graph data has high query accuracy. (4) In keeping the data availability of graph, how to keep the reachability between nodes in the process of anonymity. In this paper, a reachability-preserving-graph anonymity algorithm (RPA algorithm) is proposed. The basic idea of the RPA algorithm is to group the nodes and to adopt the greedy strategy for anonymity, so as to reduce the loss of reachability information in the anonymous process. In order to improve the efficiency of the performance of the RPA algorithm, it is proposed to use the reachable interval to effectively evaluate the anonymous loss caused by the edge adding operation; secondly, by constructing the candidate neighbor index, the anonymity process of the RPA to each node is accelerated. Through a large number of experimental analysis, the anonymous map generated by the RPA algorithm maintains the inter-node reachability, so that the anonymous graph has good data availability in the aspect of reachability query. and (5) realizing the social network data security release prototype demonstration system SNSPDEMO. The SNSPDEMO system can carry out security detection on the social network for different privacy leakage types, and visually display the information of the node and the side with the privacy leakage through the graphical interface; The SNSPDEMO system integrates the social network privacy protection technology in this paper, so as to generate a secure social network diagram that provides the corresponding privacy protection, and display the graph modification operation made by the system through the graphic interface display system, and compare the difference between the original and the safety diagram. In conclusion, based on the potential threats and challenges in the social network privacy protection, this paper studies the key technologies of the social network privacy protection, such as the node privacy protection, the privacy protection of the sensitive relation, the privacy protection of the sensitive attribute values, the maintenance of the security drawing availability, etc., so as to provide a foundation for providing more comprehensive and perfect protection for social network privacy information.
【学位授予单位】：东北大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP393.08

【相似文献】