分布式网络环境下基于行为时空特征的可信评估方法研究

发布时间：2019-01-22 13:54

【摘要】：互联网作为一种分布式网络,难以使用传统集中式网络环境下集中式的信任管理机制,从而导致网络实体在交互过程中存在诸如欺诈等恶意行为,这使得系统的有效性和可用性很难保证。显然,传统集中式信任管理机制已经不再适用于大规模分布式网络环境,动态信任管理的概念由此而来,信任管理机制借助于参与网络交互的各方实体间进行动态的信任评价,以完成网络实体的可信评估。本课题针对当前以控制为中心的用户、设备静态授信的可信评估模式中存在的问题,通过对分布式网络中实体行为时空特征因素的可信评估方法研究,并结合已有信任相关领域的研究成果,构建了基于实体行为时空特征的可信评估模型,结合分布式网络的实际特征,提高可信评估的精确性、客观性和抗毁性。论文的主要研究内容如下:(1)提出了一种分布式可信计算实体的选择方法,在已有信任管理方法以及云模型的基础上,利用了云模型定量到定性表示的特征,对分布式网络中候选可信计算实体的可信情况进行全面的评估,从而在候选实体中选出可靠的可信计算实体。(2)提出一种基于行为时空特征的信任评估方法,在信任值计算过程中运用了主观逻辑和基于推荐信任的聚类算法,对目标实体信任进行评估时综合考虑了直接信任信息、推荐信任信息和间接信任信息。除此之外,此模块也建立了对实体恶意行为的惩罚机制以及考虑到了信任的时间衰减特性。(3)提出了一种基于时间滑动窗口的信任评估方法,考虑了信任的时间衰减特性与“快降慢升”的建立信任更新机制。通过对滑动窗口的初始化,窗口滑动的条件、过期信任和历史信任记录在滑动窗口内的位置的确定以及窗口内容的替换与更新,有效地控制了信任随时间的自衰减性,保证了用户行为信任评估的可信性和可扩展性。
[Abstract]:As a distributed network, the Internet is difficult to use centralized trust management mechanism in the traditional centralized network environment, which leads to malicious behavior such as fraud in the interaction process of network entities. This makes it difficult to guarantee the effectiveness and availability of the system. Obviously, the traditional centralized trust management mechanism is no longer suitable for large-scale distributed network environment, so the concept of dynamic trust management comes into being, and trust management mechanism carries out dynamic trust evaluation with the help of the entities involved in network interaction. To complete the trusted evaluation of network entities. This paper aims at the problems existing in the current trust evaluation model of users and devices with static credit in the center of control, and studies the method of trusted evaluation of temporal and spatial characteristics of entity behavior in distributed networks. Combined with the existing research results in trust related fields, a trust evaluation model based on the temporal and spatial characteristics of entity behavior is constructed, and the accuracy, objectivity and survivability of trust assessment are improved by combining the actual characteristics of distributed networks. The main contents of this paper are as follows: (1) A method of selecting distributed trusted computing entities is proposed. Based on the existing trust management methods and cloud models, the features of quantitative to qualitative representation of cloud models are used. The trustworthiness of candidate trusted computing entities in distributed networks is evaluated comprehensively, and a reliable trusted computing entity is selected from the candidate entities. (2) A trust assessment method based on spatio-temporal characteristics of behavior is proposed. The subjective logic and the clustering algorithm based on recommendation trust are used in the calculation of trust value. The direct trust information, the recommendation trust information and the indirect trust information are considered in the evaluation of the trust of the target entity. In addition, this module also establishes the punishment mechanism for entity malicious behavior and takes into account the time attenuation of trust. (3) A trust evaluation method based on time-sliding window is proposed. The time decay characteristic of trust and the mechanism of trust building update are considered. By initializing the sliding window, the condition of sliding window, the location of expired trust and historical trust record in the sliding window, and the replacement and updating of the content of the window, the self-decay of trust with time is effectively controlled. It ensures the credibility and extensibility of user behavior trust evaluation.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.0

【相似文献】