Research On Intrusion Detection Methods For Industrial Internet Based On Convolutional Neural Networks

Now-a-days,Industrial Internet has not only been widely used in many industrial scenarios such as intelligent manufacturing,gas and water conservancy,but also become a key contributing factor to promote the construction of information infrastructure.Invasion threats to the Industrial Internet are increaseing with the rapid development of intelligent manufacturing.The computing power of Industrial Internet nodes varies greatly and there are many devices running different network protocols.Traditional intrusion detection methods are not intelligent enough to deal with the information security threats faced by the increasingly developing Industrial Internet.Motivated from the characteristics and security requirements of the Industrial Internet,this thesis combined the ability to discover data features and the characteristics of strong generalization of deep learning,then studied intelligent Industrial Internet intrusion detection methods based on convolutional neural networks.The main contents are as follows:1.Aiming at the problem of poor detection effect of traditional intrusion detection methods and the need to extract feature manually,an intelligent intrusion detection method based on Res-CNN is proposed.The industrial data is standardized and normalized into the traffic characteristic maps for strengthening relative relationship.Res-CNN is constructed to build an intelligent intrusion detection model.This method enhances the feature extraction ability by superimposing the residual structures and introducing group convolution,which improves the detection accuracy while avoiding manual feature extraction.The simulation test results show that Industrial Internet intrusion detection method based on the Res-CNN can achieve 98.61% accuracy in the gas pipeline data set,and the F1 value reaches 98.11%.2.Based on the the data characteristics of Industrial Internet,capsule network is introduced.The various entity attributes in the traffic characteristic maps are extracted by increasing the information dimension carried by the neurons.At the same time,in view of the problem that the capsule network has a large number of parameters and is difficult to be distributed in Industrial Internet,residual structure is integrated to improve the capsule network,and an intrusion detection method based on capsule network is proposed.This method improves the detection efficiency by introducing a residual structure to provide high quality input to the main capsule layer,and further improves the detection accuracy.The industrial network data set is used to test the effectiveness of the method.The accuracy and F1 value of the Industrial Internet intrusion detection method based on capsule network in the gas pipeline data set test reached 99.01% and 98.66% respectively.The advantages of this method on the water data set are more obvious than the comparison model.In short,the Industrial Internet intrusion detection method based on convolutional neural network can effectively improve indicators such as detection accuracy while ensuring intelligence.The intrusion detection method based on the capsule network performs better in indicators such as false positive rate and false alarm rate while the intrusion detection method based on Res-CNN has higher detection efficiency.The two detection methods can better meet the needs of Industrial Internet.