Abstract: The intrusion of the nearest neighbor routing nodes in the deep wireless sensor combination network has the characteristic of fast load variation,and it is difficult to effectively identify the types of attacks and abnormal network behavior.Therefore,this paper proposeda network nearest neighbor instrusion detection algorithm based on convolution filtering.Network traffic is collected in deep wireless sensor combination network,and network intrusion signal model is constructed.Energy density and attack strength of network intrusion signal are analyzed in terms of time and frequency,and blind source filtering and abnormal characteristic extraction of network information are achieved by constructing an adaptive convolution filter.Joint time-frequency analysis method is used to estimate the spectrum parameters of network intrusion feature neighbor information,and intrusion detection of wireless sensor network is done according to the abnormal distribution of spectrum features.Simulation results show that this method has high accuracy for network intrusion detection,has high recognition ability and generalization ability for the unknown network traffic sample sequence,and is superior to HHT detection method and energy management method.

Key words: Adaptive, Convolution filtering, Detection, Intrusion, Network