Research On Trust Model For Dynamic Malicious Behavior In P2P Network

Recently, as a new technology in internet, P2P networks are developing quickly. It has improved the efficiency of sharing data and information among the users. It is widely used in area such as file sharing and distributed storing. However, because of the autonomy, anonymity and dynamic property, the P2P networks face lots of safety problem. At present, dynamic malicious behavior has appeared in the networks. That is to say that the malicious nodes or group will not behave maliciously all the time. Instead, they interchange with other nodes normally at first in order to achieve highly trust value and then behave maliciously.Since the origin of these safety problems are malicious nodes, which means the malicious nodes should be suppressed at first. At present, the trust model is widely used to suppress the malicious nodes. This paper studied how to build up a model in the traditional trust model. In the traditional model, node and node will interact and then evaluate each other. In this way, we can compute the evaluations to get the trust value of every node. Nevertheless, because of the existence of dynamic malicious behavior, the trust value get from the traditional trust model may not be true. This paper studied deeply on the property of the dynamic malicious behavior and came up with a new trust model, DMBTrust. This paper introduced the mechanism of the suppression of the malicious behavior, which was that incorporated the process of the evaluations of the nodes into the computation of the trust value and simultaneously monitoring the evaluations of the nodes. Further, this paper punished false evaluations by the power and exponential function we elaborately formulate. Because of the quick attenuation of the power and exponential function, the trust value of the nodes gets by cheating will decrease fast for the malicious behavior. This paper rewarded the objective evaluation of the nodes to incentive the normal behavior. Meanwhile, every node is equal, so there is no node with privilege. Therefore, the safety problem in the traditional trust models which bringing by the imitation of the malicious node could be eliminated.At last, this paper implemented a prototype system based on DMBTrust Model and tested and verified the scene which the trust model can be applied in, and compared DMBTrust Model with the EigenTrust Model by simulation experiment, then analyzed the efficiency of the suppression of malicious nodes in the two model and then analyzed their results of the simulation carefully. The result showed that the model build in this paper could adapt better to the P2P networks which existing malicious behavior, and had obvious suppression to the malicious nodes.