节点文献

一种使服务请求通过防火墙的网格信任模型研究

Research of a Grid Trust Model Enables Service Requirements Pass Through Firewall

【作者】 王婧

【导师】 杨志敏;

【作者基本信息】 山东大学 , 计算机应用技术, 2009, 硕士

【摘要】 网格是继万维网之后出现的一种新型网络计算平台,目的是为用户提供一种全面共享资源的基础设施。但因其大规模、分布异构和动态等特性使得网格计算坏境非常复杂,提出了比传统网络环境更高更广泛的安全需求。安全问题是网格计算技术的关键问题,而信任问题是人们在解决网格安全问题中所必须面对的核心问题。本文从基本的背景知识入手,总结了网格环境的特点,分析了网格的安全需求,并指出信任机制在网格安全中的重要性,进而提出一种比较综合的网格信任模型。今天的动态跨域网格计算的一个主要障碍就是防火墙的存在,防火墙不会一下子消失,它在域间执行边界访问规则。因此,网格安全模型必须考虑防火墙的存在,而且提供一种能够穿透防火墙而不破坏本地防火墙控制策略的机制。解决该问题的一种方法是通过服务绑定技术来使网格服务请求无障碍穿越防火墙等系统设置的屏障,但是绑定的安全性基于相关协议和消息格式的安全特性,而指定Web服务访问的安全需求是基于参与端点的相关策略集合定义和实现的。针对上述问题,本文对现有信任模型及防火墙技术进行研究,提出一种可以使请求通过防火墙的网格信任模型。该模型具有诸多优点:首先是不同的域可以有不同的安全策略;其次是模型考虑了交易上下文,实体的历史数据动态影响对其信任度的测量;第三是与防火墙的兼容,又不破坏本地防火墙控制策略。网格项目的共同特点是它们都跨越多个管理域而且都要处理特殊的信任和访问权限的协商过程。因此,本文在信任模型的设计中采用了基于域的信任模型的思想。按域划分网格成员,将成员之间的信任关系分为域内信任关系和域间信任关系,设置不同的策略来处理这两种不同的信任关系。因为各自然组织具有统一的安全策略的传统设计,不同的域根据自然组织来划分;域间的信任通过为每个域设置一个域代理来建立。防火墙的通过问题存在于新成员加入网格和网格成员退出网格阶段。新成员加入阶段,本文采用域代理考察新成员身份设置成员信任值之后通知域间信任表中的异域域代理的策略,相应域代理与防火墙联动,更新设置,使来自新成员的服务请求可以跨越域边界;网格成员退出网格,与其有过交易历史的异域代理再次与防火墙联动,取消退出成员跨越域边界的权利。与防火墙联动是本论文的一部分重要工作,它通过引入代理发送XML格式的消息驱动防火墙。本文另一部分重要工作是对模型中信任关系的表示、量化和更新都给出了具体的策略,并且给出了信任决择算法。在信任值的度量上,借鉴基于主观逻辑信任模型的思想,综合了身份信任和行为信任,能更接近现实地对实体的信任进行度量。最后,使用网格模拟器GridSim对论文提出的模型进行了仿真,并且对其性能进行分析。实验结果表明,使服务请求通过防火墙的网格信任模型是可行的并且在性能上优于传统的基于域的网格信任模型。

【Abstract】 The Grid is a new kind of network computing platform arisen after the World Wild Web. Its purpose is to provide an infrastructure for the users to share resources comprehensively. But since it’s large-scaled, distributed, heterogeneous, dynamic and as for many other characteristics, it makes the environment of grid computing exceptionally complex. Therefore, the requirement of security in grid environment is much higher than that in traditional network environment. Security plays a key role in grid computing technique, while trust is the core problem people must face to in resolving grid security problems. Based on some basic background knowledge, this thesis summarized the characteristics of grid environment, analyzed the security requirements for Grid, and then pointed out the importance of trust mechanism in grid security, and finally proposed a relatively integrative grid trust model.Nowadays, one main barrier of the dynamic and across-domain grid computing is the existence of firewall. Firewall can not disappear all at once, and it performs the border access rules among domains. Therefore, grid security model must consider the existence of firewall, and must provide a kind of mechanism that can make requirements pass through the firewall without breaking the local control strategy of the firewall. One solution to this problem is service binding technique, but the security of binding is based on the related protocols’ and messages’ security property, and the security requirement for appointed Web Service access is defined and implemented based on relative policy set of participated points. In connection with the problems above, we researched the existent trust models and the firewall technique, then proposed a grid trust model which can make the service requirements pass through firewall. The model has many advantages: Firstly, there are different security policies for different domains. Secondly, the model considers the transaction context. the historical data of entity influences the measurement of trust value dynamically. Lastly, the trust model is compatible with the firewall and does not break the firewall’s local control policies. Grid programs have some common characters: They all span more than one domain; they all need to deal with particular trust and negotiate process of accessing authorities. Therefore, this thesis adopted the idea of the trust model based on domains in designing our trust model. Grid members were divided according to the domains they belonged to. And the trusting relationship was divided into intra-domain trusting relationship and extra-domain trusting relationship with different process strategies. Because of the common security strategies in natural organizations, we divided domains according to the natural organizations and set up an agency for each domain to build extra-trusting relationship. The problem about crossing firewall consists in the step of new member adding to Grid and the step of old member quitting Grid. During the step of new member adding to Grid, the agency inspects new member’s identity and sets Trust Value for it, and then informs other agencies which are in the trust table (DOTT). The agencies received notification link from firewalls to update settings to make the service request of new member across domain boundary. During the step of old member quitting Grid, other agencies which have transaction with it link with firewall once more to cancel corresponding right of the member. Linking with firewall is an important part of our work, and it is done by sending XML messages. Another important part of our work in this thesis is providing specific policies for trust expression, quantification and renewal. About trust quantification, we borrowed ideas from the trust model based on subjective logic and integrated identifiable trust and behavioral trust more realistically.At last, we used grid simulator GridSim to simulate our trust model and to analyze its performance. The experiment result shows that this trust model is workable and outgoes the performance of traditional grid trust models based on domains.

【关键词】 网格安全信任模型防火墙
【Key words】 GridSecurityTrust ModelFirewall
  • 【网络出版投稿人】 山东大学
  • 【网络出版年期】2010年 05期
节点文献中: