关键词: SPIT, 反馈评判, 共轭Dirichlet分布, 加权朴素贝叶斯, 融合

Abstract: To resolve the SPIT hidden trouble in the ALL-IP fusion network, a detection/prevention method based on feedback judgement is proposed. Introducing the participation of end users and combining the trust with reputation mechanism, either the direct or indirect information feedback can be utilized simply and effectively without loss. The improved inference algorithm for trustworthiness and reputability fully embodies the distribution characteristics of SPIT behavior and reflects the weighting relation between different influencing factors on judged result. In addition, the incremental learning algorithm provides the real-time trustworthiness and reputability and the fusion of them adjusts their individual role in feedback judgement dynamically. Experiment and analysis results showed the high accuracy and sensitivity of the proposed method, by which the SPIT can be detected and prevented efficiently.