Research On Quantitative Trust Detecting Method And Model For Nodes In WSNS

Wireless sensor networks have huge development in both military and civilian areas and has become the hottest spot of research. But in wireless sensor networks, faulty nodes or malicious nodes may led to abnormal running of the whole network when cooperating with other nodes. We call the nodes above as dishonest nodes which can make faulty ex-ecution and result during data sensing, transmission and processing. Real-time detection for dishonest nodes has important application value. In this dissertation, we build three quantitative trust detecting models to detect dishonest nodes.First of all, we propose a node trust detection model based on faulty tolerant data fu-sion. Classify extra sensing data according to aggregation value, then distinguish event nodes and dishonest nodes from abnormal nodes through level selection. Experience val-ue is calculated based on the result of data fusion with applying strategy of fault tolerant. Social trust value is calculated iteratively according to neighbor nodes. It quantize nodes’ trust value by computing comprehensive trust based on experience trust and social trust. The two models can detect current dishonest nodes.Secondly, a quantitative trust model based on attributes is proposed. Focusing on data trust, we build a quantitative trust model based on single attribute.Through the cal-culation of transitive trust between non adjacent nodes, isolate outlier and detect current dishonest nodes. Focusing on comprehensive trust, the effectiveness of the communica-tion and energy are included in trust properties. Combining with data trust, we propose a light-weight quantitative trust model based on multi-attributes to detect dishonest nodes in one period.Finally, we propose a quantitative trust model based on private factors and interactive factors to detect dishonest nodes in multi-periods. Meanwhile, models and methods for trust routing are proposed. The attributes in multi-attributes quantitative trust model are classified as private factors and interaction factors. Private factors reflect nodes’ current and historical collection behavior. Interactive factors focus on nodes’ current interactive behavior with other nodes and the relationship of peer nodes. Trust graph is created with the private trust as its vertex weight and interaction trust as its edge weight. A comprehensive trust model based on the trust graph is proposed to detect continuously dishonest node. This trust model can be used to select trust routing.In this dissertation, we build simulation experiments based on OMNeT to verify the rationality and validity of the models Results show that our trust models have high-er performance of fault detection rate, lower fault detection error and communication consumption than similar models. The routing models have higher reliability and lower energy consumption than similar models.