Research On Anti-eavesdropping And Intrusion Detection Method In WSNs

In recent years,with the rapid development of sensor technology,there are more and more application scenarios of Wireless Sensor Networks(WSNs)such as health care,environmental monitoring,commercial activities and national defense security by deploying ubiquitous low-cost,high-performance and plug-and-play wireless sensor nodes.As the main role of WSNs is to collect and transmit data,the data collected by WSNs must be transmitted securely,reliably,and timely to the decision makers.Since the nodes use wireless channel,which is vulnerable to eavesdropping and attacks,the security problem of WSNs has become increasingly important.With the limited resources of the wireless sensor node,the security measures applied to the traditional wireless networks are no longer applicable in WSN,hence a dedicated WSNs security system is urgently needed.The purpose of this thesis is to study lightweight and high performance security system for WSNs.WSNs may face data attacks and attacks of network structures,routes and corrupt data.These attacks cause disclosure of important information,network functions damage and data corruption or even denial of service.Focussing on the above attacks,this paper explores data encryption mechanism before transmission,data authentication during transmission,and attack detection of received data stream at the receiving end.The primary work of this thesis is as follows:Firstly,to deal with the eavesdropping or tampering of data during the transmission,a security mechanism based on perturbed compressed sensing is proposed.It involves security measures such as data compression,encryption and authentication with low power consumption,low computational complexity and low storage requirement.To this end,this thesis introduces encryption perturbation and authentication perturbation,which provides reliable data encryption,authentication and integrity guarantee without additional communication overhead.Extensive experiments based on real-life dataset are conducted for evaluation;the results demonstrate the effectiveness in multi-hop WSNs.Secondly,at the final receiving point of the data(sink node),considering the statistical properties of data change in an unpredictable way over time during the attack period,a concept drift based ensemble incremental learning approach(CDIL)is proposed in this thesis for intrusion detection.It monitors the data steam transmitted to the sink node from the entire wireless sensor network and can handle new types of attacks that cannot be identified by current model.This is achieved by updating the intrusion detection model based on the data stream containing the new attack.For classification model,a dual ensemble classifier including a master and a standby ensemble classifiers is employed to achieve soft classifier update.In this way,the learning model can adapt to the changing input network status data in real time.Extensive experiments are conducted on a special WSNs intrusion detection data set and prove the effectiveness and real-time performance of the proposed CDIL.