Security Decision-making Mechanism, Based On The Probability Of Node Trust Game

Depending on the various features of Peer to Peer (P2P) network, such as self-organization; openness and anonymity, P2P network have been widely applied in multiple domains of the Internet. Meanwhile, the continuously growing P2P security problems pose a serious threat to the development of the P2P technology. To avoid such a threat and to stay in control of the spreading of the malicious behavior in P2P network, researchers have already proposed many effective ways. Currently, trust mechanism which is based on reputation concept is the main approach and has been applied to most of the P2P file-share systems .This dissertation proposed a new decision making mechanism which can predict and analyze the P2P node behavior to help the P2P network nodes make the approximately optimal decision during the procedure of the P2P interaction.First of all, this dissertation provides an overview on the current state of research on trust mechanism, including the difficulties trust mechanism has encountered, the classification of trust model and system. While most of the research efforts carried on this filed are focused on evaluating the reputation value by iterating history interaction score among P2P nodes. So far, how to"accurately"define the reputation value to represent the node past behavior is still remain unsolved .Therefore, this dissertation focuses on this central problem by introducing"node benefit"concept into whole reputation system .Secondly, as the Bayesian Network(BN) is regarded as the effective way to make optimal decisions, this dissertation proposes a new method based on BN to predict the node behavior. The algorithm extracts every node and other nodes with which it had interacted to construct a simple Bayesian Network. Then it employs the Bayes'rule to calculate the probability of nodes'behavior. Besides, the algorithm make the trust model sensitively detect malicious behavior and keep robust when malicious attack occurs. Experiment shows that then new algorithm has a better performance when compared to non-BN algorithm.Thirdly, this dissertation proposes the assumption that the node interaction procedure is considered as a long term, continuous process rather than a discrete one. Thus, the node behavior which aims at maximizing node's long term benefit become the significant factor to determine whether the node interaction should continue or not. For that purpose, the second half of the dissertation discuss the long term repeated game theory to prove the existence of Nash Equilibrium (NE) in this particular situation. At last, according to the result of NE, the node behavior which is used to maximize node's long term benefit can be accurately predicted to help both two sides of the interaction make the approximately optimal decision.