Improving Intrusion Detection Precision via Multi-Classification Algorithm Examination

Network Intrusion Detection is crucial for safeguarding digital infrastructure against unauthorized access and malicious activities. By continuously monitoring network traffic, intrusion detection systems identify and analyze suspicious behavior, such as unauthorized access attempts, malware activity, or anomalous data transfers. This proactive approach helps in swiftly detecting and mitigating security breaches, minimizing potential damage to systems, data, and network integrity. It is possible to configure machine learning methods within the Network Intrusion Detection System (NIDS), which makes it easier and faster to identify network threats. In this study, the experiment utilizes Kyoto’s 2015 May Day1 dataset. Different machine learning classification methods such as KNN, AdaBoostClassifier, Naïve Bayes, XGBClasifier, and DecisionTreeClassifier are applied to the supplied dataset for training-size 80% and test-size 20%, as well as for training-size 70% and test-size 60% and for training-size 60% and test-size 40%. The results of this study will surely help other practitioners to choose an optimal machine learning algorithm for a Network Intrusion Detection system. This study’s ultimate objective is to give classification algorithms ability for Network Intrusion Detection, enabling practitioners to choose the best strategies for situations that are similar to their own. n this paper, the best accuracy was achieved through the following algorithms, LGBM, Decision Tree Classifier, GradientBClassifier, these algorithms gives the accuracy of 100%. The accuracy of proposed method compared with state-of-art and proposed method accuracy is the best accuracy among all state-of-art algorithms. The source code following this research is available at https://github.com/varun-v-1410/Kyoto_2015_May_day1.