A Research On Key Techniques Of Trust Management For WSN Based On Collective Intelligence

In recent years, Wireless Sensor Network(WSN) has been playing an important role in many fields, such as military affair, environmental protection, medical treatment, etc., and has become a key component in sense layer of Internet of Things. Because of the opening environment of WSN applications, adversaries can capture and reprogram nodes to laugh internal attacks easily. Security policies based on encryption can hardly deal with these attacks. Trust management can establish a monitoring mechanism within the network, through which nodes can evaluate neighbors’ behavior and WSN can build a supervision mechanism based on reputations of nodes to establish security policy. Trust management in WSN ensures that nodes can collect data safely, reliably, efficiently even if internal attacks appear.Since WSN is usually non-center and self-organized, with dynamic topology and weak devices, it’s hard to achieve intelligent trust management in WSN. By drawing lessons from social animals which show intelligence through collaboration, trust management based on collective intelligence has been proposed to deal with this issue. A great number of nodes with weak devices cooperate with each other and show intelligence in trust management so as to protect the security of WSN, improve the capability of intrusion tolerance and proof the safety, reliability, efficiency of data collection. Approaches proposed in this paper can overcome the contradiction between the sensitivity of intrusion detection and intrusion tolerance in traditional trust management frameworks, deal with the issue of frequent abnormal data caused by hardware malfunction and proof the balanced transmission of nodes, configurable precision of data aggregation. The main work and contributions of this paper are as follows:1. An intelligent model of evaluation and representation in trust management of WSNs is proposed. According to the principle of artificial intelligence with uncertainty, we design Lightweight Cloud Model based on Cloud Model, and propose the uncertainty evaluation of direct trust, indirect trust, and recommendation trust in WSNs to realize the representation of trust relationships between nodes intelligently. Simulation results show that this method is not only tolerant of abnormal conditions, but also sensitive to multiple attacks.2. A real-time abnormal data filtering approach based on collective trust is proposed in WSN. Through this approach, quantitative data can be converted to qualitative knowledge to describe the characters of data, and suspicious data can be detected by comparison with the basic characters of data. On the theoretical basis of the spatio-temporal correlation of data in WSNs, collective trust of suspicious data can be computed based on the comparisons of qualitative knowledge of neighbors so as to detect and filter abnormal data. Simulation results show that this method can not only detect and filter the outliner in-time, but also cost less in transmission and computation.3. An in-group scale-independency data representation approach is proposed. In the research of prediction in WSN, the rule of data prediction based on a date set is found below:the absolute error of prediction is proportional to the dispersion degree of the data set. Based on this finding, theory analysis, formula derivation, mathematical proofs are made, and several theorems and corollaries are found. Furthermore, a scale-independent data representation in dataset is proposed. On the basis of this representation,"similar vector" and "inner error" are put forward and applied to collaborative filtering and WSN data collection field. The validity of these methods has been proved by simulation experiment.4. An approach on data colloction in WSN on the basis of in-group error control is raised. The data collection of WSN usually needs following characteristic:transmission balance, configurable precision, low cost, and future knowledge independency. According to the boundedness of inner error, an in-group precision control model is proposed. In this model, every node adapts its threshold from recent historical data to achieve balanced transmissions; the adaptive threshold can control the error of aggregations in real-time. Precision configurable transmission balanced data collection approach is proposed in cluster-based WSNs based on this model. Theoretical analysis and experiment results show that nodes can save energy effectively through this approach. Compared with Bernoulli sampling-based aggregation, MAEs of continuous queries of sum aggregation and average aggregation is quite small by this approach. Furthermore, this approach needs no future knowledge so that it is usable and adaptive in many WSN applications.