ENHANCING NETWORK INTRUSION DETECTION AND CLASSIFICATION BY USING HYBRID MACHINE LEARNING APPROACHES

The present era is the modern technology evolving era for cybersecurity. It boons a dynamic battlefield for cyber security concerns for security experts. Network intrusions have become a major concern in cyberspace for compromising security. Traditional methods like manual rules, blacklists, and whitelists are insufficient for detecting modern intrusions. While machine learning approaches for intrusion detection have emerged, many suffer from low accuracy. However, recent advances in machine learning algorithms show promise for improving intrusion detection and classification. To address the limitations of current methods, this work proposes a hybrid machine learning approach for intrusion detection and classification. The approach utilizes seven classifiers including decision tree, random forest, naïve Bayes, ADA, XGB, KNN, and logistic regression. The model is evaluated on the CICIDS2017 dataset using training and testing splits. The classifiers achieve accuracy rates of 0.99 for decision tree, 0.96 for random forest, 0.85 for naïve Bayes, 0.97 for ADA, 0.96 for XGB, 0.98 for KNN, and 0.91 for logistic regression. The decision tree classifier demonstrates the highest accuracy of 0.99, owing to its effective parametric function evaluation and ability to minimize misclassification errors. The proposed hybrid approach aims to advance network intrusion detection and classification capabilities beyond current techniques.