可信网络中一种敏感鲁棒的信任计算模型

Sensitive and robust model of trust computation in trusted networks

提出一种敏感、鲁棒的信任度计算模型。在缺乏直接交易证据时,基于推荐信任、信誉等间接信任度估计信任度初值,随交易的建立和累积与直接交易证据共同作出信任评价,并最终过渡至直接信任度计算。采用动态平衡系数,兼顾信任度初值估值和直接交易证据小样本风险,逐步降低信任度初值估值风险并将其限制在最初的有限次交易以内;同时增加交易证据对信任度的贡献。在直接信任度计算中,采用双滑动窗口机制,即稳定性窗口N和敏感性窗口M,使信任度保持相对稳定又对交易满意度动态敏感。采用动态权值法赋予明显偏离信任度均值的节点信任度较小权值,提高模型的抗攻击能力。给出了直接信任度稳定性窗口N、敏感性窗口M、稳定性—敏感性平衡系数的建议值。仿真结果验证了本方法的有效性。

This paper proposed a sensitive and robust trust computation model.At the stage for lack of direct transaction proofs,estimated an initial trust value based on recommendation trust or reputation.With the construction of new transactions and their accumulation,calculated trust by combining of the initial value and new transaction proofs,then transferred to the direct trust computation model,at last.It adopted a balancing coefficient to weaken risks in estimating initial trust values,and restricted them within limited initial transactions.At the same time,strengthened the contribution of proofs from direct transactions,step by step.In the computation of direct trust,a double window mechanism,which included a stability window N and a sensibility window M were adopted respectively to make trust relative stable but sensitive.It integrated node reputation by weighted trust degrees from other nodes.To strengthen the abilities of anti-attacking,trust values obviously deviating from mean were dynamically set with small weights.Proposed values of stability window N,sensitiveness window M,and stability-sensitiveness balance coefficient of direct trust model.Simulation results prove the effectiveness of the proposed model.