Research On WSNs Intrusion Detection Technology Based On Deep Learning

With the rapid development of Wireless Sensor Networks(WSNs)in various fields,their security issues have become the focus of attention.Due to the limited energy of sensor nodes in WSNs,small network data transmission throughput,poor real-time intrusion detection,and low accuracy,to achieve WSNs intrusion detection,we must first ensure low network energy consumption and long life cycle.Therefore,specific analysis is needed to design an intrusion detection model that conforms to the characteristics of WSNs.Firstly,in view of the wide deployment area and large amount of data collected by WSNs,considering the node characteristics,network topology and communication mode of WSNs.Cluster-based WSNs intrusion detection scheme is designed,which can do fast and efficient data collection,easy management and wide deployment range.Secondly,in order to increase the energy consumption of nodes in the process of WSNs intrusion detection,combined with the energy-limited characteristics of sensor nodes,an improved C-FCM routing protocol is proposed to save energy and ensure the effective implementation of intrusion detection algorithms.Based on the traditional C-FCM protocol,the cluster head number selection,cluster grouping and cluster head selection algorithm and data transmission strategy are improved.The simulation is carried out with Python software,and compared with LEACH routing protocol,the analysis is carried out in five aspects: clustering,number of surviving nodes,energy consumption of cluster head nodes,total energy consumption of the network and data transmission.The simulation results prove that this paper proposes that the routing protocol saves the energy of the WSNs node,increases the amount of data transmission,and extends the life cycle of the network.Then,in order to improve the intrusion detection efficiency of WSNs,combined with the deep learning algorithm,a clustering node data compression and anomaly detection algorithm based on SAESM is proposed for the problem of poor real-time detection and small data throughput.By combining the SAE compression algorithm in deep learning with the SVM binary classifier at the neural network level,the cluster head node data compression is realized,and anomaly detection is realized.Simulations were performed using Python software,and data compression,anomaly detection,and data throughput were analyzed.By comparing with other algorithms,it is proved that the algorithm of this paper effectively increases the throughput of the network and can achieve fast local intrusion response in the sensing area.Finally,in order to accurately determine the attack type and identify the new attack behavior,aiming at the low accuracy and low real-time performance of the existing advanced intrusion detection algorithm,combined with the deep learning algorithm,an advanced detection algorithm based on SLSTM is proposed.The SAE algorithm is effectively combined with the LSTM algorithm to fully exploit the temporal correlation between attack data.The Python is used to simulate the hidden layer unit number,accuracy and loss function,and compare with other algorithms.It proves that the detection rate of the SLSTM advanced detection algorithm proposed in this paper is as high as 97.83%,which can judge the attack more accurately.The algorithm was transplanted to the hardware platform for real-time testing.The detection time was 33.051 seconds.Compared with other algorithms,it proved that the proposed intrusion detection algorithm has better real-time performance.