云计算环境下动态多元信任模型与方法研究

Cloud computing has been generally attracting the attention of society both in the academia and the industry as it provides an on-demand computing, high efficiency, scalability, economy and high availability services. However, security and trust problem resulted from the dynamic and uncertainty characteristics are perceived as primary obstacles to its further development. Around cloud computing dynamic multivariate trust modeling, algorithm and methodology, this project is mainly carried out the following research: (1) The overall topology architecture has been designed to provide a reasonable and efficient dynamic multivariate trust model for cloud Computing. Specifically, the proposed multivariate trust model incorporates three aspects: direct trust, recommendation trust and compensation trust. The discount attenuation trust relationship, consensus rule trust relationship and time sliding window mechanisms are interacted to calculate the trading trust values. (2) A compensation trust model has been constructed to develop multivariate trust model in cloud Computing environment. The measure and computing methods of compensation trust value have been put forward to address the different failure problem of cloud services and the principle and mechanism of collaborative work with direct and recommendation trusts. (3) A dynamic evolution and incentive trust model, which is based on trading profit incentive polymerization, has been developed to provide the dynamic calculation and update algorithm for trading income value. The mechanisms of service push and autonomic matching have been also carried out on our dynamic multivariate trust model. (4) A hybrid cloud platform and its testing system have been deployed to test and evaluate our dynamic multivariate trust model in cloud computing environment.

云计算作为一种具备高效性、可扩展性、经济性和高可用性的按需计算新技术，近年来得到了社会的普遍关注。然而其动态性和不确定性等特征带来的云服务信任问题是其进一步发展面临的主要挑战。围绕云计算动态多元信任建模和方法设计，本项目主要开展如下研究：（1）设计合理高效的云计算动态多元信任总体模型和拓扑架构，提出多元信任（直接信任｜推荐信任｜补偿信任）模型，研究折扣衰减信任关系、合意运算信任关系、时间窗口滑动综合作用的动态信任机制和交易信任度计算方法。（2）构建云计算多元信任管理的补偿信任模型，研究不同云服务失败情况下补偿信任值的度量和计算方法，以及引入补偿信任机制后云计算多元信任模型的协同工作原理与机制。（3）研究基于交易收益奖惩聚合的动态演化信任激励模型与交易收益值的动态计算和更新算法，并给出最佳信任服务推送和自主匹配机制。（4）设计并实现一个支持动态多元信任模型的混合云平台及其测试评价系统。

当今，云计算、物联网等技术飞速发展，各种云服务模式层出不穷，如电商平台、即时通讯软件、社交媒体等。云服务应用具备开放性和动态性特征，给用户带来了非常便捷的使用体验和较高质量的服务形式。不过，也导致了比普通网络应用环境更为严重的信任和安全问题，制约了云服务模式的进一步推广和发展。如何通过合理的信任模型和隐私保护机制，更好地刻画云服务开放性和动态性等特征，是研究人员关注的热点问题。. 围绕着云计算环境下信任模型与方法研究这一主要问题，项目主要开展了三项研究内容：云环境中的隐私信任保护模型研究、云环境中的社交网络行为分析研究、云计算的基本数据处理算法研究。在隐私信任保护模型方面，项目重点研究了差分隐私模型，分别提出了基于直方图排序、基于集合覆盖、基于多层抽样以及基于多维边缘表的差分隐私模型；在社交网络行为分析方面，项目提出了公众邻居袭击防御、重要节点估计、DDoS攻击检测、高效位置服务及近邻查询等算法；在基本数据处理算法方面，项目研究了可扩展性聚类、分布式哈希、本地性信任推荐、基于遗传算法的任务调度等内容。. 按照上述各项基本研究要点，项目依照计划循序渐进地开展，总体执行情况良好，取得了一些有意义的成果，发表学术论文26篇，其中SCI检索论文10篇，包括IEEE/ACM Trans期刊2篇；EI检索论文16篇，包括本领域知名国际会议7篇，如ICME、ICA3PP、TrustCom等；参加27人次国际学术会议，做大会分组报告14人次；培养博士研究生4名，硕士研究生18名。