Improving Intrusion Detection System using Feature Extraction

To enhance our daily activities, globalization and the interoperability of computing systems are increasingly prevalent. However, this also exposes vulnerabilities that exceed human control. As a result, communication exchanges must now incorporate cybersecurity measures. To ensure secure communication, it’s essential to enhance security protocols to address evolving threats and effectively combat these challenges. Given that network behaviour is constantly evolving, it is crucial to thoroughly assess various datasets using various methodologies. A network-based intrusion detection system has been suggested in this paper. On a publicly accessible benchmark dataset CICIDS 2017, a study of the performance of several machine learning classifiers and Deep Neural Networks (DNN) for both binary and multi-class classification was conducted. The outcomes of the deep learning and machine learning algorithms have been compared to those of the reference paper. The experimental results showed that, when compared to the Deep learning method, our machine learning models outperform it in both binary and multi-class classification.