Research And Implementation On Trust Model Based On User Behavior

With the rapid popularization of Internet, the development of the Internetand related technologies is rapid, and the amount of Internet users is becoming agreat one and diversification of network services are the main challenges facedby today’s CDN. In order to save costs and provide better quality of service, theintegration between P2P and CDN technology is becoming the focus of thecurrent research. P2P hybrid CDN can well solve the failure of single point,performance bottlenecks,conserve bandwidth and so on. Security risks in P2Pare introduced a new threats into hybrid CDN. In order to provide more reliableservice in the new hybrid network system, by studying and learning from theexisting trust model, we proposed classification trust mechanism based on theinteraction history to solve the security problem in the network.First of all, proxy cache server response mode in this trust model will bedivided into two categories, the response mode is high-speed response for userswho have a high trust value, but for the average user, the response mode iscommon mode response after the review of trust rules. The proxy cache serverrefused to add these nodes that with a lower trust value into the autonomy P2Pdomain. The response of proxy cache server is varying according to the level ofthe user, we believe that there will be more nodes sharing resources andproviding the real recommended information in order to get a better server’sservice, and at the same time keeping down untrusted file-sharing behavior.Secondly, taking into account of that the construction and destruction of trustvalue is unsymmetrical, so we using the linear rise and exponential decline in thetrust value calculation process in order to raise the loss of the malicious nodes.This model also introduces a time adjustment factor to ensure that the mostrecent interactive evaluation reference value. By the transfer of the credibility ofthe recommended trust value of the recommended node and the update rules ofthe credibility of the recommend node to obtain a more realistic indirect trustvalue.Finally, We conducted a simulation experiment, by comparison with other typical trust models, we get that our trust model has a better security, it can resisteffectively the malicious attacks of all kinds of malicious nodes.