The Research On Malicious Node Detection For Wireless Sensor Networks

Wireless sensor network is disposed in a open, collaborative and random environment. Because of the fragility link among nodes,nodes exposed in physical circumstance, varied topology dynamically and shortage of identity authentication, there are a lot of security problem. Especially, inner malicious nodes are hard to be defensed,so it is significant to study to detect malicious nodes exactly.According to the analyses of the attack threat of WSNs and existing security strategies, this thesis emphasize particularly on research and betterment of malicious-node detection.For nodes with unconspicuous attack character, this thesis take a modified reputation evaluating model to optimize the evaluating of reputation, the model not only need direct reputation but also take into account indirect reputation information from other nodes ,this thesis take reliability detection strategy to learn from other nodes' experience and avoid defaming and commending. This model punishes selfish nodes effectively. Simulation result shows this model can improve speed and veracity in detecting malicious nodes.For blind-letter malicious nodes,this thesis optimize neighbor watch scheme to detect whether nodes transmit packets really and offset the shortage of watch dog.For nodes with conspicuous attack character, this thesis takes different strategy according to different feature and takes hop-by-hop verification based responding nodes to detect blackhole nodes.According to transmission character of sensor network,this thesis extend existing routing protocol and utilize transmiting list to detect malicious nodes forging packets.This means is proved effective by simulation, and it is a light-load and low energy-consume scheme suitable for resource-restricted WSNs.Finally,the work done in this thesis has been given and the future work has been brought forward.