Abstract: Because of its openness, diversity and complexity, the network, trustiness faces the severe challenge. In order to guarantee the credibility of the whole system, it is very essential to establish a valid management model based on trusted relationship among nodes. At the same time, the new emerging computing model based on independent cooperation is considered, which, will influence the development trend of network node behavior. Based on the trustiness of network node behavior, this paper abstractly makes the network system as the manager monitor agent (MMA) with hierarchical structure based on trustworthy networks and agent technique. Based on all those above, the paper mainly studies the logical components, function and strategy in MMA model. Through analyzing various factors influencing the node behavior, the paper describes a new method and strategy which can monitor and measure the node behavior on the expectancy behavior character for trusted computing of the node. According to the manifestation of node behavior and historical information, the trust-value of node behavior is adjusted and predicted, and the trusted-interval is defined to realize the organization and management of the trusting relationship among nodes. The paper then establishesa dynamic trust evaluation model based on node behavior characters. Finally, the experimental results shows the effectiveness and feasibility of the model.