Research On Some Key Techniques Of Malicious Nodes Detection In Wireless Sensor Networks

By collecting plenty of various sensor data and fusion processing, wirelesssensor networks can put out a series of meaningful information. With the developmentof science and technology, the application of wireless sensor networks covers moreand more areas, such as, ocean environment detection, wildlife observation, buildingsecurity, seismic monitoring, highway intelligent traffic, forest fire alarm, waterquality monitoring, intelligent healthcare and so on.However,resulting from thecharacteristics of wireless sensor networks, the security problem is more and moreobvious: firstly, most sensor nodesare limited in energy,communication range andprocessing capability.Secondly, differing from traditional network, wireless sensornetwork is usually deployed in outdoors andunmannedenvironment, security threatenof physical attack will come about. In the end, wireless sensor networks is easy to beaffected by physical environment and anthropic factor, resulting in some securityproblems. These factors of networks make traditional network security technologyand countermeasures can’t be applied directly and make the nodes more easier to becaptured.As long as the enemy can easily capture the nodes,the key data can be readand the whole networks may be destroyed. After analyzing some usual type of attack,some security models and malicious node detection algorithms are studied. The mainworksare as followed：(1) As for the small node quantities and planar network structure, test data packagewill be used to inspect malicious node which may start selective forward attack andtime attack. Firstly, test data package will be sent to in-network nodes by base station.And they are requested to forward to neighbor nodes, and suspect them; Secondly,nodes will vote the suspicious of neighbor nodes by comparing the package sent bybase station and neighbor nodes; At last, base station will select out the maliciousnodes by nodes voting. The experiment results show that MNDRVM has higherdetection rate and lower false detection rate of malicious nodes.MNDRVM algorithmis more optimized compared withmultivariateclassification algorithm. The falsedetection rate of MNDRVM algorithm is lower than WTE and multivariateclassification algorithm.(2) As to the large node quantities and hierarchical network structure, for theshortages that WTE and WCF algorithm limit1-hop communication in cluster, and fusion accuracy is affected by malicious nodes.Firstly, high reputation value filteringmechanism is added in fusion algorithm to detect malicious sensor nodes.Secondly,data package counting method is proposed to inspect the malicious forwarding nodes.The experiment results show that, MNDSDF algorithm is more close to ideal valuecompared with WTE. MNDSDF has higher detection rate and lower false detectionrate compared with WCF, and it is more suitable to most of routing protocols.