An Efficient Approach Using Graphical Neural Network for Network Intrusion Detection System
Shailendra Pratap Singh, Bansal Abhay, Akshat Kumar, Prakhar Agrawal, Druhan Singh, Vivashwan Ghosh · 2024
This investigation is centered on addressing the imperative for a System on Intrusion Detection capable of accurately discerning anomalies and attacks within a system. As the frequency and severity of network intrusion incidents continue to rise, the necessity for a robust IDS becomes increasingly pronounced. To contribute to this endeavor, we have devised a classification model specifically designed for network intrusion detection, leveraging advanced techniques.In today’s interconnected landscape, network security is of utmost importance. Network intrusion attacks pose formidable challenges for all computer networks and can incur substantial costs for manufacturers. Consequently, the significance of a robust and dependable Intrusion Detection System with the ability to identifying and attacks within the system is steadily increasing. In this research, we primarily focused on developing a multinomial classification model for network intrusion system. We trained our model on a dataset including various attack types, such as R2L, PROBE, DOS, and U2R, alongside typical attacks. To achieve it, we explored ML methods such as decision trees, k-nearest neighbors, and random forest classifiers. We measured each algorithm’s performance based on metrics like accuracy, precision, recall, and F1-score. Notably, our research and numbers indicate that the GNN algorithm outperformed others, achieving an accuracy rate of $91 \%$.