Research On Security Technology Based On Trust Models For Wireless Sensor Network

As an important component of Internet of Things, the research on Wireless Sensor Network(WSN) has been further developed in recent years. It has been widely used in national defense and military, agricultural production, health care, environmental monitoring, logistics tracking and other various aspects. Different from traditional networks, the computing, storage and energy resources of wireless sensor network are limited. And it is usually deployed in the unattended and open environment. All these features make wireless sensor network extremely vulnerable to damage and serious security threats. Traditional security technologies based on encryption and authentication can only resist external attacks, and can not effectively solve internal attacks caused by node capture. In addition, the encryption and authentication algorithms are complex and require additional computational costs, which are not suitable for wireless sensor networks with limited resources. Therefore, the design of an effective security mechanism for wireless sensor networks has become a hot issue in wireless sensor networks. In this context, the research on trust management mechanism in wireless sensor networks arises at the historic moment due to its lower computational complexity and stronger ability to resist internal attacks.The dissertation takes the trust management mechanism of wireless sensor networks as the research object, and studies trust management model which is suitable for wireless sensor networks by considering the characteristics of WSNs and multiple challenges existing in trust management mechanism. In general, the main contents of the dissertation conclude the following three aspects:(1) Considering the fuzziness and uncertainty of trust, a multi-factor subjective logic trust evaluation model is proposed in the dissertation. Considering the security demands, the main task as well as the contradictions between the safety and energy consumption of the network, typical trust factors are selected from three aspects, i.e. communication characteristics, data properties and physical properties to calculate the node direct trust. Furthermore, use the concept of fuzzy sets for fuzzy partition, and take the obtained fuzzy degree function as the basic belief degree function in the D-S evidence theory for the fusion of direct trust and indirect trust. Taking into account the security threats in the process of trust information transition, the identity-based encryption mechanism is introduced to ensure the security of the trust information in the transmission process.(2) The existing trust models ignore the abnormal behaviors caused by the failure of the network itself. In order to solve this problem, an improved trust model based on Bayes theory is proposed. Taking into account the abnormal behavior of network caused by non-intrusion factor, the Bayesian formula is modified by introducing the abnormal attenuation factor. And the sliding window and the adaptive forgetting factor are used to update trust. In order to reduce the influence of the energy consumption and malicious feedback, the dissertation judges the confidence level of the direct trust to determine whether it can be trusted as the comprehensive trust. If the confidence level is lower, then calculate the indirect trust to obtain the comprehensive trust. The information entropy theory is used to assign weights, which can overcome the limitations caused by the subjective weight assignment and strengthen the adaptability of the model itself. Simulation results show that the proposed model has a higher detection rate and a lower false detection rate, which greatly reduces the energy consumption of the network.(3) Study the trust evaluation models for clustered wireless sensor network. Under the premise of meeting the network security demand, a lightweight and reliable trust model for clustered WSN is proposed to reduce the overheads and improve the reliability of the network. Based on the cluster topology, a lightweight trust evaluation scheme is introduced, in which the trust evaluation scheme is selected according to the roles of nodes in the clustered wireless sensor network. Furthermore, use the concept of fuzzy nearness degree in the fuzzy theory to measure the credibility of the third-party recommendation, and define the adaptive weight approach for trust fusion, which can reduce the influence of malicious feedback, and also weaken the influence of collusion attack. Theoretical and experimental results show that, compared with other classical trust models for clustered wireless sensor network, the proposed model requires less memory and communication overheads, and has good fault tolerance and robustness.