A Dynamic Trust Model Based On Task Classification In P2P Networks

With the development of Internet, the application of distribute networks becomes more and more widely. P2P networks as a classical structure of distribute networks applies widely either. Because of the open and anonymity nature of P2P networks, there are some problems about the security. How to solve the problems about security of the P2P networks becomes a popular topic. A good safety mechanism can make the P2P networks working safely and normally. To set up a safety, trustworthy trust management model is a best way to solve the security problems of P2P networks.There are many security problems in P2P networks, including virus transition problems, query flooding problems and malicious behavior problems. Malicious behavior problem is one of them which is most diverse and difficult to prevent, including fraud, conspiracy slander hidden, Free-Riding, etc. any combination of them can form a new malicious behavior. Presently a dynamic trust management is a common and effective method to solve these security problems. Dynamic trust management improves the accuracy of the model through a comprehensive assessment of multiple factors, and designs a security trust model for a variety of malicious attacks against conduct to improve the trust management mechanism for practical applications. This makes the trust model of multi-attribute more adaptive, and be able to solve a variety of application environments of trust management. Meanwhile the credible decision-making dynamic is one of the dynamic trust management problems which is highlighted, and can solve decision-making independently and shortfall response capability problems of traditional network security system. The research of the relationship of trust and risk, trust and cooperation through the combination of trust and incentive mechanism can solve the problems of distributed application for autonomous agents.This paper does some research and analysis about some trust models concludes the definition of trust and some natures referred, and include the classes of trust relations and the expression of trust. We classified the trust model which is existed according to the calculation of trust to six categories: online model, probability and statistics based on trust model, subjective perceptions based on trust model, the trust model based on fuzzy logic, flow model and behavior of software base on trust model. Trust has some properties including asymmetry, multi-mapping, delivery, predictability, reliability, measurability, dynamic etc. While this paper also analyses the integrity measurement mechanism advanced by Trusted Computing Group (TCG) and another trust measurement mechanism, and compared them to find a way to set up a trust model. According to the security problems in P2P networks, we set up a dynamic trust model based on task-level classified.The trust model we contribute considers that the tasks which the nodes in P2P networks interactive would be not same, the risk they can bare is not the same, we classify the task of some levels, and set up a dynamic trust threshold. The higher the task level is, the higher the trust threshold is. We compute the trust with both direct trust and indirect trust to defend the malice active. At the last of the paper, we design a experiment environment, which implement the model by simulation, and do some experiments and analysis for the model. The experiment analyzes the feasibility of the model from three malicious behaviors including simply malicious nodes, malicious evaluation node, the node re-entry. At the same time, we do an experiment for the task classified of the model to show feasibility and advantage.