Trust Computing Method In Social Network

With the development of Internet, more and more Internet users use social networks to record andshare personal status, share and exchange transactions, Upon obtaining the required services throughsocial networking, which several opponents can provide it, the users need to evaluate each serviceprovider is a trust value, according to the trust value to make a choice.Social networks as a particular class of applications, most of them employing P2P topology structure,the main work of this paper is the research of trust computation method in the social network with theP2P network topology. Firstly, summarizes the current P2P network trust calculation models, methodsand some specific algorithms, analyses existing security problems in current trust computing.Combined with the characteristics of P2P network topology and the social network, the trustcomputation model is divided into two categories, one is dynamic trust model based on transactionfeedback, the other is static trust model based on relationship and its depth and interest similarity.Secondly, In order to solve two existing problems, the time dependence influence is not consideredwell and the recommendation trust value based on individual is not accurate, we put forward twospecific algorithms, moreover considering time attenuation factor, stimulation and punishment factoretc. to reduce the complexity of the algorithm, improve its computational accuracy and anti-attackcapability.In view of existed trust algorithms cannot very good representation of time correlation, this paper isbased on the Markov chain methods, proposes a steady-state probability dynamic prediction algorithmto describe trust development trend. Weighting the transition probability and the steady-stateprobability to improve the accuracy of the calculation of trust degree, at the same time, instead equaltime segment based time attenuation factor of the number of transaction time attenuation factor toreduce the complexity of the algorithm, imtroduce penalty excitation factor to improve its anti-attackability. The analysis and experiments shows that the algorithm can reduce the computation complexity,improve the algorithm accuracy and the ability of anti attack.With the growing number of groups based community in social network, this paper proposed grouprecommendation model with reputation. Most existing models adopt individual recommendation tocalculate strangers’ trust, but individual recommendation may be not accurate due to the individualcognitive limitations, Reputation is more objective and accurate than individual recommendation.Moreover, As the size of the community are generally not large, the calculate of the reputation is of not complexity. Finally, proposed community association degree to reduce the collusionrecommended risk, greatly improving the model’s anti-attack ability. Experiments provide theselection standards of community association degree and promote algorithm’s accuracy.