Research On Key Technologies Of WSN’s Malicious Nodes Detection Based On Credit Mechanism

With the rapid development of science and technology, the purpose foremploying a wireless sensor network extended gradually from single usage formilitary to each area regarding national economy and the people’s livelihood.By the same time, security issue are becoming more and more serious. Tracingit to its causes, primarily, it is caused by the inherent structure of wireless sensornetwork, it is more easier to attack if the network is more opening. Secondarily,because the low costing hardware is relatively poor quality, it can not directlyapply on various security defense technology which usually using in traditionalnetwork. Besides, wireless sensor networks is often deployed in the unattendedarea, this will not only lead to the failure of high-intensity operations and safetycertification its own energy supply shortage led to wireless sensor nodes, butalso caused external attacks because of imperfect security relatively.In recent years, a wireless sensor network malicious nodes detectionmethod based on the credibility mechanism is widely used in wireless sensornetwork security solutions. It identify malicious nodes through quantify nodesbehavior, although people can get good results usually, but still exist kinds ofdeficiency. The contents of this thesis is for solving problems that based oncredibility model algorithm, like ’over-reliance on artificial experience’,’poor identification of hidden malicious nodes for attack behavior’ and ’irreconcilablebetween normal nodes identification and malicious nodes misjudgement rate ’,in order to improve the typical credibility model. At the same time, also for thespecial circumstances and requirements of wireless sensor network security, willoptimize clustering algorithms that used in it. The main contribution of thisthesis are:A wireless sensor network node credibility detection algorithm based ontiming sequence(Time Sequence Based WSNs Nodes Reputation Detection,TSB-WSNNRD) is proposed in this thesis. By introducing concept "Timingcredit collection", expanding a long time cycle from a single time point of nodesreputation value which based on typical credibility model algorithm, therefore,the wireless sensor network nodes could judge reputation value which made bynodes behavior more objectively. Meanwhile, introducing clustering algorithminto wireless sensor network’s reputation measure, with the help of clusteranalysis method, it not only can correctly identify the hidden malicious nodes innetwork, but also make sure the balance between normal nodes identificationand malicious nodes misjudgement rate in algorithm. Ultimately, in order tomade algorithm obtain practical value, combine with the own characteristics ofwireless sensor network security, the typical K-medoids algorithm is optimized,and an improved K-medoids algorithm based on center(CLB-K-medoids) isproposed, which is greatly improved the algorithm efficiency, to a certain extent,solved problems like high time complexity and sensitive initial value that are existing in K-medoids algorithm, make sure it is competent for the work ofwireless sensor network security testing. Last but not least, through a series ofsimulation experiments shows that the performance of this algorithm inidentifying malicious nodes in wireless sensor network have a high accuracyand stability, and can be able to accomplish required tasks with in shortestpossible time.