面向海量信息的可信服务预测方法研究

There is a new challenge for trust model according to the characters of enormous scale and high-speed generation for huge information. Security system of traditional solution methods called “build strong fire walls, hit patches” method can not satisfy security needs. Lots of new distributed services and applications imperatively require a credible network environment that ensures the security and trust availability of huge network source. Here, trusted prediction schemes play an important role in the field of trust. Firstly, many objective and subjective factors impacting on systems’ trust attribute are deeply analyzed. Based on Web mining algorithm the discovery, combination and renewal mechanism of trust paths are presented for huge information envionment, which will make trust evolution rules be rational and valid. The behavior characters of single entity and collusive organization are studied under huge data environment and recognition methods of collusive organization are presented, which is based on clustering algorithm. It can help to make the punishment strategy. Combining with data mining methods, the novel trusted prediction schemes and algorithms are designed, which depended on objective and subjective computing tools. Finally, these new algorithims and protocols will be verified on experiment platform and an application system about services composition, which is based on trust model, will be constituted.

海量信息的巨大规模和高速产生的特点为可信任技术提出了新的挑战。传统的“筑高墙，打补丁”的安全体系解决方案已无法适应需求，许多新的分布式应用迫切要求搭建一个可信型的网络互联环境，以确保海量资源的安全共享和可信利用。其中，可信预测扮演着越来越重要的角色。本课题首先深入分析影响系统可信性的多种主观和客观因素，基于Web挖掘算法，研究海量信息环境下信任路径的发现、融合和更新机制，以确保信任演化规则的合理性和有效性；探讨海量数据中单个恶意实体和共谋团体的行为特征，基于聚类算法设计共谋团体的识别方法并确立惩罚策略；结合数据挖掘方法，设计主客观计算工具相融合的可信预测方案和相关算法；最后，借助仿真平台验证所提的算法和协议，并建立一个可信驱动的服务组合应用示范系统。

本课题首先深入分析影响系统可信性的多种主观和客观因素，基于Web挖掘算法，研究海量信息环境下信任路径的发现、融合和更新机制，以确保信任演化规则的合理性和有效性；探讨海量数据中单个恶意实体和共谋团体的行为特征，基于聚类算法设计共谋团体的识别方法并确立惩罚策略；结合数据挖掘方法，设计主客观计算工具相融合的可信预测方案和相关算法；最后，建立了一个可信驱动的服务组合应用示范系统。