Chinese Journal of Computers   Full Text
Title	Adaptive Dynamic Trust Measurement and Prediction Model Based on Behavior Monitoring
Authors	LI Xiao-Yong GUI Xiao-Lin MAO Qian LENG Dong-Qi
Address	(Department of Computer Science and Technology, Xi’an Jiaotong University, Xi’an 710049)
Year	2009
Issue	No.4(664—674)
Abstract &   Background	Abstract In the large-scale distributed systems, trust relationship model is one of the most complex concepts in social relationships, and it also is an abstract psychological cognitive process, involving assumptions, expectations, behavior and the environment, and other factors. So, it is very difficult to quantify and predict trust relationship accurately. In this paper, rough set theory and information entropy theory are combined and applied to the study of distributed dynamic trust measurement and prediction model based on behavior data. The new model works through analysis monitored behavior data by sensors directly, changes the traditional modeling thoughts, brakes away from the fetter of various subjective assumptions in traditional modeling methods, and overcomes the problem of inadequate handling capacity for multi-source behavior data in the traditional trust model. Simulating results shows that the new model can accurately implement trust measurement and prediction process between entities in open and complex distributed environment, and has a better scalable capacity of behavior data. Keywords information security; dynamic trust model; rough set; information entropy Background With the widespread applications of large-scale open environments, such as Grid computing, Ubiquitous computing, P2P computing, Ad hoc networks, etc., the technology of dynamic trust management has become a significant requirement from a network security’s point of view, and trust evaluating and predicting mechanism has become a determining factor for any given service’s success. But the dynamic nature of trust creates the biggest challenge in measuring trust value and predicting trust relationship amongst peers. In recent years, many of state-of-the-art trust models have been proposed, and some of them are very innovative and elaborate, but most of the studies still have some limitations: (1) Many current trust models use simple or one-sided trust decision factors to quantify and predict trustworthiness of service providers or requesters, which may lead to inaccurate or unfair outcome of trust decision. The authors think that when trust relationship between peers cannot be fairly defined, it is unstable, and difficult to manage and predict. (2) In many of previous studies, the subjective assigning method to weights of trust decision factors cannot reflect trust decision scientific and reasonable, and may lead to misjudgment of trust decision result. Focusing on these problems, in this paper, rough set theory and information entropy theory, are combined and applied to the study of distributed dynamic trust measurement and prediction model based on behavior data. Firstly, a new trusted decision-making method based on historical evidences window is proposed, which not only can reduce the risk and improve system efficiency, but also can solve trust measurement and prediction problem when the direct behavior evidences are insufficient. Then, this paper focuses on trust measurement and prediction model based on behavior evidences: (1) Using the concept of rough set knowledge expression system, trust decision table is set up based on time-stamp; (2) Using fuzzy aggregation methods, the new model categories the history evidence records (domain) composed of multi-source monitoring data under different confidence level and obtains relevant knowledge. (3) It uses information entropy theory to determine the classification weight of trust attributes (indicators), and finally implements fusion computing of overall trust degree. The new model works through analysis monitored behavior data by sensors directly, changes the traditional modeling thoughts, brakes away from the fetter of various subjective assumptions in traditional modeling methods, and overcomes the problem of inadequate handling capacity for multi-source behavior data in the traditional trust model. Simulating results shows that the new model can accurately implement trust measurement and prediction process between entities in open and complex distributed environment, and has a better scalable capacity of behavior data. This work is supported by the National Nature Science Foundation of China (No.60873071); the National High Technology Research and Development Program (863 Program) of China (No.2008AA01Z410); Program for New Century Excellent Talents in University of China (NCET No.05-0829); Scientific and Technological Project in Shaanxi Province, China (No.2007K04-05).