Artificial Immune Based Intrusion Detection In Wireless Sensor Networks

The wireless sensor network (WSN), as a new information acquisition technology, hasgradually become a new advanced research field which attracts many researchers and is fusedby multi-discipline. The characteristis of WSN, including that huge number of sensor nodes,deploying in unguarded area, multi-hop communication easy to be wiretaped and disturbed,distribution and self-organization, make the security research of WSN face great challenge.This paper mainly studies on WSN intrusion detection technology, analyzes the main existingthreats and attacks in WSN, chooses and extracts the corresponding feature, and proposes ahierarchical hybrid intrusion detection algorithm. The artificial immune system which is thelatest achievements in artificial intelligence is introduced in WSN intrusion detection, and twointrusion detection algorithms based on immune algorithm are proposed.(1) WSN hierarchical hybrid intrusion detection algorithm. Construct a two-levelintrusion detection model. Before detecting intrusion, the principal component analysis isadopted to reduce the feature dimension so as to reduce the data storage and computationspace. In the cluster level, the transductive confidence machine for K-nearest neighborsalgorithm is used in anomaly detection in the sensor nodes, then the support vector machinewhich is optimized by particle swarm optimization algorithm is used in misuse detection incluster heads to classify the anomaly data present by the sensor nodes. In the base station level,the anomaly detection and misuse detection are combined to deal with the monitoring datapresent by cluster heads with high detection rate and low false positive rate. The simulationresults show that the proposed detection algorithm can achieve high accuracy even in the caseof small samples.(2) WSN intrusion detection algorithm based on improved V-detector. Taking fulladvantage of the characteristics of abundant resources, select the training samples andgenerate and optimize the detectors in base stationton. In ordinary nodes, the sensor node isresponsible for data collection and features selection. In detection nodes, the dimension of thefeatures selected by ordinary nodes is reduced, and the samples are successively detected withmemory detector set and mature detector set. The V-detector algorithm is supplemented andimproved in five aspects to adapt the resource constrained WSN. They are the featureselection and processing of samples, the training sample screening, the detector generationrules, the detector optimization algorithm and the detection rules.(3) WSN anomaly detection algorithm based on rough set theory and improved dendriticcell algorithm (DCA). The anomaly detection framework is constructed based on the principleof biological immune system. The attribute reduction method in rough set theory is adopted toreduce the signal dimension to reduce storage and computation space. The input signalselection mechanism is set at the same time. To improve the based on danger theory, thedendritic cells (DC) capacity of lymph node is set and DC update mechanism is introduced toreduce the storage space and guarantee the freshness of DC. Change the migration thresholdvalue from fixed value to regional to reduce energy consumption of communication. Modifythe antigen abnormal standards from static to dynamic to realize dynamics detection networkanomalies in real-time.The simulation results show that: Principal component analysis method and rough set attribute reduction method can achieve good effect of dimension reduction; Hierarchicalhybrid detection algorithm can reduce the false positive rate and false negative rate at thesame time under the condition of small sample; Detection algorithm based on V-detector canreduce storage and computation space, improve detection accuracy, and can quickly respondto secondary attack; Detection algorithm based on improved dendritic cell algorithm canreal-time detect network anomalies and has high detection accuracy.