基于隐Markov过程的网络信任评估模型

为了能快速精确地刻画实体行为的高动态性，提出了一种基于连续时间隐Markov过程的信任评估模型。不同于离散时间隐Markov信任模型，该模型充分考虑到信任的时间相关性，结合交互记录之间的时间间隔，将实体信任评估问题归结为连续时间隐Markov过程的学习问题。进而利用改进的和声搜索算法，给出了求解隐Markov过程最佳参数的算法，该算法有效地保证了全局搜索空间，能够获得更好的解。在此基础上，利用已有交互结果序列和最优参数组，对实体的信任度进行预测。仿真实验表明，该模型能够快速地反映出实体行为的动态性，具有较高的精确度，且能抵抗部分恶意攻击。

A Dynamic Trust Evaluation Model Based on Optimized Hidden Markov Process

In order to depict high dynamic of entity behavior quickly and accurately, a trust evaluation model based on continuous-time hidden Markov process is proposed in this paper. Different from the trust models on discrete-time hidden Markov chain, this model fully considers time dependence of trust, combines time intervals between the interactions and makes the trust evaluation problem boil down to the learning problem of continuous-time hidden Markov process. Then an algorithm for solving the optimal parameters of hidden Markov process is given with the improved harmony algorithm, which can effectively guarantee the global search space and achieve a better solution. On this basis, the trust degree can be predicted using the existing interaction sequences and optimal parameters. Simulation results show that the model is able to quickly reflect the dynamic of entity behavior, has high accuracy and resists the malicious attacks.