【作者】 傅翀；

【导师】 秦志光；

【作者基本信息】 电子科技大学 ， 信息与通信工程， 2011， 博士

【摘要】 近年来,互联网上的各种隐私泄露及侵犯事件层出不穷,隐私保护也因此成为信息安全研究的一个重要课题。匿名通信是隐私保护技术的重要组成部分,因而受到广泛地关注。时间攻击是对匿名通信威胁极强的一种攻击方式,对其进行抵御的关键在于如何在匿名保护和通信性能之间取得合理的平衡。基于此,本文对抗时间攻击技术进行了研究,取得的创新成果包括如下4个方面:1.提出了一种基于水印破坏的防御机制,用于抵御主动式时间攻击中最典型的水印攻击。对于非交互式网络流或在实时性要求较高的环境中,该机制采用时延规范化技术来破坏攻击者嵌入的水印;对于交互式网络流或在实时性要求较低的环境中,该机制则利用时隙干扰技术将水印破坏。在真实网络流数据集上的实验证明该机制可以在有效保证通信性能的前提下消除水印,达到防御的目的。2.建立了一种基于流集合分割的掩饰添加流变换机制,并基于此提出了动态组调度和基于层次加权近邻传播聚类的关联链路填充两种流变换方法。前一种方法将流集合按时域特征分簇并进一步分组后再进行流变换,后一种方法则引入层次加权近邻传播算法,利用流聚类技术对关联链路填充流变换算法进行了改进。实验表明两种方法都能够显著地降低流变换使用的掩饰流量,同时也证明了基于流集合分割的掩饰添加流变换机制能够有效地提高系统通信性能。3.提出了一种基于多路复用的关联链路填充流变换机制。该机制将关联链路填充算法部署到分布式匿名系统的节点上,建立了基于多路复用的关联链路填充流变换方法,并构建了基于卡尔曼滤波的多路复用关联链路填充流状态预测模型,然后在其基础上提出了一种流带宽动态调度算法。仿真实验表明该机制在带宽有限的节点上进行流变换时可以有效降低掩饰流量,并且能够保障正常网络流的带宽需求。4.提出了一种MIX型匿名系统中的抗被动式时间攻击等级服务模型,并在Tor协议基础上对模型进行了实现。该模型将匿名通信划分为多个等级,每个等级代表了不同的匿名性和通信性能的组合,然后定义了一组服务等级函数,根据用户所选等级使用相应函数来为其选择通信链路上的路由节点。该模型在入口节点上实施基于多路复用的关联链路填充流变换,并在后续节点上依照不同等级对掩饰流进行相应程度的逐级消减,从而提高通信性能。仿真实验证明了该等级服务模型的可行性和有效性。更多还原

【Abstract】 Invasion of privacy, which is a serious problem on the Internet in recent years, has made the privacy protection playing an important role in the research on information security. Anonymous communication is one of the most important subjects of private protection technique and thus has attracted a lot of attention. Timing attack is a serious attacking method on anonymous communication. The main point to defend against it is to make a reasonable balance between anonymity and performance of communication. On this basis, the research on the defensive technology against timing attack is carried out in this dissertation. Four innovative contributions of this dissertation are enumerated as follows:1. A watermark destroying mechanism is proposed to defend against timing watermarking attack, which is the typical active timing attack method. On one hand, for non-interactive flows or the case where the real-time requirement is high, the proposed mechanism destroys the watermark by applying the delay normalization technology. On the other hand, for interactive flows or the case where the real-time requirement is low, the proposed mechanism destroys the watermark by implementing time slot disturbance technique. Experiments on the flow data sets of real network prove that this proposed mechanism can remove the watermark while guaranteeing the communication performance of the anonymous system.2. A mechanism of cover traffic-adding flow transformation based on flows set division is proposed. On this basis, a dynamical group scheduling method and a Dependent Link Padding method based on Hierarchical Weighted Affinity Propagation algorithm are proposed. The first methods divides the flows set into clusters and further divides each cluster into groups, and then deploy flow transformation on each group. The second one improves the Dependent Link Padding algorithm by combining it with Hierarchical Weighted Affinity Propagation clustering algorithm. Experiments shows that the above two methods can reduce the volume of cover traffic when deploying flow transformation. Accordingly, it also proves that the proposed mechanism can improve the communication performance effectively. 3. A mechanism of Multiplexing Based Dependent Link Padding flow transformation is proposed. The mechanism deploys Dependent Link Padding algorithm at a node in distributed anonymous system, designs a Multiplexing Based Dependent Link Padding method, and builds a flow status prediction model of Multiplexing Based Dependent Link Padding based on Kalman filter. On this basis, a dynamical flow bandwidth scheduling algorithm is proposed. Simulation shows that the proposed mechanism can not only reduce the volume of cover traffic efficiently when deploying flow transformation at the node with limited bandwidth, but also guarantee the bandwidth requirement of normal flows.4. A multi-level service model for defending passive timing attack in MIX anonymous system is proposed, and the model is implemented based on Tor protocol. The proposed model divides anonymous communication into several levels, each of which represents a certain combination of anonymity and communication performance, and then defines a family of service level functions so that the routers of the ciruit can be selected by using the function corresponding to the level selected by the user. In the model, Multiplexing Based Dependent Link Padding is deployed at the entry node of the flow path, and the cover traffic is reduced gradually at the subsequent nodes according to the selected level. Therefore, the communication performance is improved. Simulation proves that the proposed model is valid and practicable.更多还原