Vice Chancellery for Research and Technology, University of Tabriz Advanced Signal Processing 2676-3397 5 1 2021 08 23 RETRACTED ARTICLE: (A model for multi-class intrusion detection with imbalanced data in the CICIDS-2017 dataset) RETRACTED ARTICLE: (A model for multi-class intrusion detection with imbalanced data in the CICIDS-2017 dataset) 105 115 14336 10.22034/jasp.2022.48285.1165 FA Mahmoud Niaei Faculty of Management and Accounting, Azad University, Research Sciences, Tehran, Iran Jafar Tanha Faculty of Electrical and Computer Engineering, University of Tabriz, Tabriz, Iran Gholamreza Shahmohammadi Faculty of Electrical and Computer Engineering, Ivan Key University, Semnan, Iran Alireza Poorebrahimi Faculty of Management and Accounting, Islamic Azad University, Karaj, Iran Journal Article 2021 10 11 <strong>This paper is retracted according to the COPE Retraction Guidelines:</strong><br /><strong>• The findings have previously been published elsewhere (<span style="text-decoration: underline;">http://pitc.jrl.police.ir/article_97273.html</span>) without proper attribution to previous sources or disclosure to the editor, permission to republish, or justification (ie, cases of redundant publication)</strong><br /><br />Today, most economic, commercial, cultural, social and governmental activities and interactions in all countries are carried out through cyberspace. Due to the inherent vulnerabilities in cyberspace, the risks of systems are increasing. Therefore, the security of networks and systems against various types of intrusion has become one of the most important challenges of the present age. In this research, a model for detecting network intrusion has been reviewed and proposed. The proposed method is a multi-class method and the dragonfly algorithm is used for feature selection and the Random forest algorithm is used for classification. For analysis, the CICIDS-2017 unbalanced data set has been used, so the balancing operation has been used. To select the method, different algorithms are tested and the best algorithm is selected. The value of accuracy in the proposed method is 0.9985. In addition, the research results have been compared with several other methods proposed by previous researchers, and this comparison shows that the proposed method were better than most of the researches presented in the article. <strong>This paper is retracted according to the COPE Retraction Guidelines:</strong><br /><strong>• The findings have previously been published elsewhere (<span style="text-decoration: underline;">http://pitc.jrl.police.ir/article_97273.html</span>) without proper attribution to previous sources or disclosure to the editor, permission to republish, or justification (ie, cases of redundant publication)</strong><br /><br />Today, most economic, commercial, cultural, social and governmental activities and interactions in all countries are carried out through cyberspace. Due to the inherent vulnerabilities in cyberspace, the risks of systems are increasing. Therefore, the security of networks and systems against various types of intrusion has become one of the most important challenges of the present age. In this research, a model for detecting network intrusion has been reviewed and proposed. The proposed method is a multi-class method and the dragonfly algorithm is used for feature selection and the Random forest algorithm is used for classification. For analysis, the CICIDS-2017 unbalanced data set has been used, so the balancing operation has been used. To select the method, different algorithms are tested and the best algorithm is selected. The value of accuracy in the proposed method is 0.9985. In addition, the research results have been compared with several other methods proposed by previous researchers, and this comparison shows that the proposed method were better than most of the researches presented in the article. https://jasp.tabrizu.ac.ir/article_14336_c504c946229582378e291ebb0c7665c5.pdf