The Design And Implementation Of Intrusion Detection Firewall Based On Machine Learning

The development of the Internet has brought many conveniences.While enjoying these conveniences,people are also experiencing the threat of increasingly fierce and complex network abnormal attacks.Traditional signature-based firewalls use rule matching for intrusion detection.They have insufficient detection capabilities for unknown threats and zero-day vulnerabilities.Intrusion detection for massive traffic and logs in a stand-alone system also faces bottlenecks in processing time,business response,and storage space.Therefore,this paper designs an intrusion detection firewall based on machine learning and deploys it on the Spark platform for implementation.The main work of this paper are as follows:(1)Aiming at the problem of the limited results of a single feature selection method,a Weight Ensemble feature selection(WEFS)method is designed,which uses the idea of integrated learning algorithm to aggregate multiple feature selection algorithms and give each feature selection algorithm at the same time Different weights are merged,and finally a selected feature subset is obtained.WEFS is conducive to further improving the generalization performance of the integrated feature selection method,selecting the most important features in the data set,and effectively improving the efficiency and performance of the intrusion detection firewall.(2)Through the analysis of the principles,advantages and disadvantages of the Deep Autoencoding Gaussian Mixture Model(DAGMM)of the machine learning algorithm,improvements and optimizations are proposed for the disadvantages of DAGMM,and the problems caused by direct splicing of joint variables are proposed.The feature weighted regularization operation is performed on the joint variables,and the network structure is adjusted for the problems of the network structure being prone to disappearing gradient and too small dimensionality reduction.and the optimal dimensionality reduction vector dimension value is determined through experiments,and an improved depth is obtained.The Improved Deep Autoencoding Gaussian Mixture Model(IDAGMM)algorithm improves detection performance and accuracy compared with the original DAGMM algorithm.(3)Design a set of intrusion detection firewall based on machine learning,and deploy it on the Spark Cluster for implementation.Combining WEFS with the improved IDAGMM algorithm to Weight Ensemble feature selection-Deep Autoencoding Gaussian Mixture Model(WEFS-DAGMM)algorithm as a machine learning intrusion detection algorithm to detect abnormal intrusions,enhance the network safety.The WEFS-DAGMM based intrusion detection firewall designed and implemented in this paper improves the feature selection method and machine learning algorithm,and deploys the machine learning algorithm on the Spark platform for intrusion detection.After experiments and system tests,it is confirmed that it has usability and accuracy and can effectively improve the performance of intrusion detection.