基于混合特征学习数学模型的WSNs入侵检测研究

Research on WSNs Intrusion Detection Based on a Mathematical Model of Hybrid Feature Learning

随着WSNs入侵方式的多样化,现有检测算法的准确性会显著下降,为此提出了基于混合特征学习数学模型的WSNs入侵检测方法。检测中先基于嗅探器定时获取传感器节点运行参数,并利用熵权法选取无线传感器网络节点混合特征因子,包括能耗率、丢包率、报文发送频率、报文接收频率,再从获取传感器节点运行参数中获得四个特征因子对应的数值。利用BP神经网络构建入侵检测数学模型,以混合特征因子数值为输入,确定WSNs入侵类型,以实现对WSNs入侵行为的精确检测。结果表明:基于强化学习的入侵检测模型,基于双向循环生成对抗网络的入侵检测模型相比,所研究模型应用下,F1-measure数值更大,由此说明该模型入侵检测更全面,检测结果更准确,模型的整体性能更好。

With the diversification of WSNs intrusion methods,the accuracy of existing detection algorithms decreases significantly,for this reason,a WSNs intrusion detection method based on a mathematical model of hybrid feature learning is proposed.In the detection,the sensor node operation parameters are first obtained based on the sniffer timing,and the entropy weight method is used to select the hybrid feature factors of the wireless sensor network node,including the energy consumption rate,the packet loss rate,the message sending frequency,and the message receiving frequency,and then the corresponding values of the four feature factors are obtained from the acquisition of the sensor node operation parameters.A BP neural network is used to construct an intrusion detection mathematical model to determine the type of WSNs intrusion using the values of the hybrid feature factors as inputs,in order to achieve accurate detection of WSNs intrusion behavior.The results show that the intrusion detection model based on reinforcement learning,and the intrusion detection model based on bidirectional recurrent generative adversarial network has a larger value of F1-measure under the application of the model under study compared to the intrusion detection model based on bidirectional recurrent generative adversarial network,which indicates that the model intrusion detection is more comprehensive,the detection results are more accurate,and the overall performance of the model is good.