Research On Intrusion Detection Classification Method Based On Convolutional Neural Network

Intrusion detection system can effectively detect attacks in the network,do timely prevention,avoid greater harm.Through the in-depth investigation of the research background and research status of convolutional neural network(CNN)and intrusion detection technology at home and abroad,it is found that the classification performance of intrusion detection technology has certain limitations in the face of massive and high-dimensional data,while CNN has certain advantages in data processing.Therefore,this paper studies the intrusion detection method based on CNN algorithm.The main contents of this paper are as follows.First of all,the problem of CNN's unsatisfactory learning performance for one-dimensional data,and the imbalance of categories in the intrusion detection data leads to the problem that it is easy to be biased to most category samples during the training model.This paper proposes the use of data oversampling and data format conversion methods,using the enhanced and converted image data to train the CNN model.The data conversion process is to expand the dimension of one-dimensional data into even-numbered column features,fill the expanded features with random numbers that conform to the normal distribution,and then convert them into two-dimensional image data.Secondly,,the current data exhibits high-dimensional and massive characteristics,and traditional machine learning algorithms cannot effectively extract the features required by the classifier,resulting in a decrease in the performance of the classifier.In response to this problem,this paper takes advantage of CNN's ability to automatically extract features,and proposes CNN intrusion detection algorithms based on image enhancement and intrusion detection algorithms based on CNN and Light Gradient Boosting Machine(Light GBM).In order to achieve the goal of improving the classification performance of the intrusion detection system and reducing the false alarm rate in the face of high-dimensional and massive data.Finally,the proposed CNN intrusion detection algorithm based on image enhancement and the intrusion detection algorithm based on CNN and Light GBM are experimentally verified and analyzed.The intrusion detection algorithm proposed in this paper has improved overall accuracy,overall accuracy,and accuracy of each category.The overall false alarm rate has been reduced,especially in the classification accuracy of small samples.The research work of this paper shows that the algorithm proposed in this paper has achieved relatively ideal results in intrusion detection experiments,and therefore has certain application value.