云计算环境下基于行为的动态信任模型研究

The clouding computing is a new model based on Web Service, in which the entities are autonomous, open and dynamic. So there is no trust relationship among them and it is difficult to select the high quality service for the customer and service composition. The behavior-based dynamic trust model is the primary solution. However most of current behavior trust models have many defects such as the dimension and credibility of the ratings, reflecting the dynamic characteristic of the trust, the storing and access of the trust data and the access control on the service requests that are from the different trust domains. Pointing to these problems, we will give the following methods: 1) a dynamic trust model based on the anomaly detection will be proposed, in which multidimensional feature vector will be used to describe the quality of service and the difference between the test vector and the sample average vector will be computed that is used to forecast the reality of the ratings; 2) a record-based dynamic model will be presented that can adjust the current trust according to statistical result of the historical behavior, so the dynamic characteristic of the trust can be reflected better; 3) a QoS-based mechanism of storing and accessing trust data will be given that not only can help the customers to search those services met the requirements of the given quality,but also is efficient and secure; 4)For the sake of resolving the authorization between the different domains, a new authorization policy based on the behavior trust will be proposed that is not only simple, but also invigorative. According to the methods mentioned above, this project can solve the defects of the current trust model in the cloud computing environment. So our practice will promote the application of the cloud services.

云计算是基于Web Service的新型计算模式，具有自治、开放和动态特性，实体间缺乏信任，服务消费者和服务组合难以选择到高QoS服务。基于行为的动态信任是解决此问题的有效方法。然而，云环境下基于行为的动态信任模型在评价维度和可信度、反映信任动态特性、信任数据存储与访问、域间访问控制等方面存在问题。基于此，本项目拟展开如下研究：1）拟提出一种基于异常检测的动态信任模型，用多维特征向量描述QoS属性，计算测试向量与样本平均向量之间距离，从而预测评价真实性；2）拟提出一个基于行为记录的动态信任模型，根据历史行为统计结果调整目前信任，能更好地反映行为信任动态特性；3）拟提出一种基于QoS属性的信任数据存储与访问机制，既能查找满足QoS需求的服务，又高效和安全；4）拟提出一个基于信任策略的域间授权机制，既简单，也有激励作用。综上所述，本项目实施将对解决云环境下信任缺失、促进云应用具有重要意义。

虽然云计算目前非常流行，但大多数企业用户还是不愿意将其核心业务部署到云上，主要原因就是信任缺失问题，表现为：在云计算环境下用户失去了数据控制权；云服务器缺乏透明性；云服务提供商给出的安全保证是不清晰的等。因此为了增强用户对云计算的信心，信任机制应运而生，成为研究热点。本项目从以下几个方面展开了研究：.首先提出了Stadam，一个新的基于异常检测和多云的SLA信任模型。在Stadam里，云用户不仅能在服务运行时检测到异常云服务，还能够在多个云服务之间实现无缝切换。实验显示，Stadam无论是在可用性、平均响应时间、平均成本，还是成功处理的请求个数方面，都较现有模型好。.接下来提出了一个适应P2P存储云环境的安全、高效和细粒度的访问控制机制（简称ACPC）。通过安全分析，ACPC是安全的；性能评价显示，ACPC在实际环境下是高效的；提出了一个适用于移动云计算环境的数据访问控制方案SL-AC。SL-AC由属性加密算法SL-CP-ABE和属性签名算法ABS-AT构成。为了减轻移动终端的计算开销，SL-AC将绝大部分计算外包至代理服务器，仅保留很少的计算在移动端；并且没有降低系统的安全性。.最后，本项目提出一种基于MapReduce 编程框架的云数据审计方案，使用代理签名技术，将用户的数据签名计算代理出去；并且并行化处理数据签名和批量审计过程。实验结果表明，本文提出的方法明显提高了批量审计的效率，增强了云存储服务的可用性，并且减轻了用户的在线负担；另外，本项目没有使用加密技术，而是基于纠栅码和拉格朗日插值法实现了云数据的可用性和机密性等。通过实验分析，提出的方法在数据上传和下载响应时间上都有所改进。.综上所述，本项目的实施能够改善云环境的安全性，增强云用户对云服务的信任，从而促进云计算的应用。