Intrusion Detection System using NSL-KDD Dataset and Decision Tree Machine Learning Algorithm

Abstract

A firewall is designed to filter traffic, block unauthorized access, and allow authorized access. Modern network hardware contains vulnerabilities that can be exploited by attackers to compromise network security. The aim of this project is to propose an efficient machine learning model to analyze the vulnerabilities in firewall logs. The project implemented decision tree classifier on a NSL-KDD dataset pulled from Kaggle ML repository. This dataset contains 127,000+ instances of vulnerabilities. In order to build a web app, the input features of the dataset were reduced from 41 to 5 within the Streamlit interface to enhance prediction speed, reduce resource consumption etc. The following performance metrics were obtained for the selected model: Accuracy: 99.31%; Precision: 98.98%; Recall: 99.54% and F1-Score: 99.26%. The inclusion of quantitative metrics, visual evaluation and a real time web application prototype that helps bridge academic exploration and industrial relevance serves as a contribution to knowledge in firewall detection system. The outcome of the research proves that all these measures of accuracy show that the decision tree model is precise, reliable and suitable for practical implementation in analysis of firewall logs in order to detect the existing vulnerabilities for system engineers. There are some distinct benefits of applying machine learning to the problem of firewall vulnerability analysis, hence, this study has been able to contribute to the field of firewall and intrusion detection system in modern hardware system.