An Adaptive Recognition Algorithm for Intrusion Behavior in Active Communication Networks Based on Improved Deep Learning
Aimed at the problem that the external parameters are sensitive and the identification accuracy is seriously affected when the parameters change greatly,an adaptive identification algorithm of active communication network intrusion behavior based on improved deep learning is designed.We normalize the active communication data,combine the convolutional neural network and BGRU,construct an improved cyclic neural network for end-to-end attack detection,optimize the activation func-tion and logistic regression classifier,and identify the intrusion behavior of active communication network stably and adaptive-ly.Experimental results show that the proposed algorithm can still maintain high recognition accuracy when the convolution kernel size and learning rate change,and the recognition results of active communication network intrusion are adaptive.
万方数据
维普
