Network Intrusion Detection Method Based on Multi-Scale CNN in Internet of Things

Abstract

Network intrusion detection is a powerful means to identify and analyze the state of the Internet of things. For the reliability requirements of the Internet of things, an intrusion detection analysis method of the Internet of things based on a deep network model is proposed. First, based on the Inception network architecture as the backbone network, this method constructs a multi-scale convolutional neural network (M-CNN) intrusion detection analysis network model. The long-term and short-term memory network models are introduced into M-CNN to enhance the local feature extraction ability of the model. At the same time, batch normalization and global average pooling layers are introduced to make the data distribution of each layer uniform, reduce the model training time, reduce the model calculation gradient, and further improve the efficiency of the network. The simulation experiment takes the KDDcup99 data set as an example, and the results show that the M-CNN intrusion detection model can achieve better results. The precision P r e and recall R e of the detection model are 93.90% and 93.59%, respectively.

Research on Adaptive 1DCNN Network Intrusion Detection Technology Based on BSGM Mixed Sampling

“…Deep-learning methods possess powerful feature extraction capabilities and can handle massive high-dimensional data [6][7][8]. Yin et al [9] proposed a multi-scale convolutional neural network called M-CNN for IoT security. This model incorporates long short-term memory networks into M-CNN to enhance local feature extraction capabilities and utilizes the Inception network architecture as the backbone network to construct a multi-scale convolutional neural network.…”