Identification and processing of network abnormal events based on network intrusion detection algorithm

With the popularity of the Internet, people's lives are becoming more and more convenient, but the network se- curity problems are also becoming increasingly serious. In order to better prevent internal or external malicious attacks and protect the network security of users, this study chose deep neural network (DNN) learning algorithm and convolutional neural network (CNN) learning algorithm as network intrusion detection algorithms and tested two algorithms under different parameters and activation functions with KDD99 data set on the MATLAB simulation platform. Moreover, the performance of the algorithms was compared with those of other clinic algorithms and deep learning algorithms. The results suggested that the recognition performance of DNN and CNN learning algorithms was different under different network parameters and activation functions. When ReLU function was used as the activation function, the recognition performance was the best. The network parameters of DNN and CNN were 122-250-520-250-5 and was 10(18)-14(22)-16 (18), respectively. The recognition performance of DNN and CNN learning algorithms were better than those of the classical algorithms, self-organizing map (SOM) and support vector machine (SVM) algorithms, but was worse than that of dynamic Bayesian network (DBN) algorithm. DNN was superior to DBN in the aspect of false alarm rate; overall, DNN algorithm was superior to DBM algorithm. © 2019 Femto Technique Co., Ltd.