The Trust Of The Wireless Sensor Network Research

In wireless sensor network (WSN) security solutions, cryptography alone is not sufficient enough to secure the whole network in terms of the distinct characteristics and misbehaviors of WSN nodes. Partially because internal compromised nodes still have valid cryptographic keys to access the other nodes within the network. In addition, the computing ability and power source available in battery-driven WSN nodes are limited and could be insufficient for processing complex security algorithms like those in MANET and P2P. Recently, reputation and trust have been introduced into WSNs to overcome the above deficiencies. The reputation metric is constructed based on the statistical properties or observation consistency of sensory data. When a sensor node produces sensory data with statistical properties that are deviated from the norm, its reputation is considered tarnished. Therefore, not only the performance and robustness of the network can be enhanced, but also trust or reputation mechanism can provide a stimulating effect for cooperation among nodes.In the reputation modeling of WSNs many literatures have proposed novel reputation indirect update methods such as reputation integration, discounting, and aging to eliminate and filter malicious and false reputation information. However, few have discussed the reputation direct update. In this dissertation, based on the analysis of problems incurred by the current WSN reputation update methods based on beta distribution and according to the negative binomial distribution in statistics, a novel reputation update method for WSNs is proposed. Results show that when a WSN node is compromised, its reputation value will decrease faster so that reputation computing power and storage overhead of the neighboring nodes can be saved accordingly, which will be more suitable for the resource constraint WSNs.Secondly, reputation update under improved certainty is put forth in this research. There exist some deficiencies in traditional certainty method for WSN reputation modeling and the algorithm of the traditional certainty method is more complex, time and energy consuming. While the improved method in this study is more general and the algorithm is simpler, which also meets the constraints of WSNs. Thirdly, as a simple application of reputation update based on negative binomial distribution, a tasked based reputation update method for WSNs is proposed. According to the characteristics of tasks, reputation need not update each time after one transaction or it can update after a fixed number of transactions. This method can set the parameters of the reputation model in consistence with the different tasks to make the reputation update mode comply with the actual task demand.