Enabling Robust Intrusion Detection in Network Traffic through an Integrated Machine Learning Framework

In the contemporary cyber landscape, intrusion detection is a critical security concern. A plethora of techniques, primarily grounded in machine learning, have been devised to address this challenge. This research presents an interconnected module framework for intrusion detection utilizing machine learning algorithms. The system’s architecture is designed for efficiency in intrusion detection tasks. The data-driven approach includes modules for Data Selection, Data Preprocessing, Data Splitting, Classification (employing Logistic Regression and Convolutional Neural Networks), and Performance Analysis. The systematic class diagram outlines each module’s role, ensuring a comprehensive approach to developing a machine learning-based intrusion detection system. Hyperparameter Tuning is employed for LR and CNN models, optimizing performance. Explainability and Interpretability are emphasized through SHAP values, enhancing transparency in decision-making. Results cover system requirements, testing strategies, and software choices, ensuring robustness. The proposed system demonstrates compatibility with hardware and software specifications. Dataset analysis, data splitting results, confusion matrices, and performance metrics showcase system accuracy. The accuracy of 97.56% is achieved using LR and 99.99% using CNN. The research provides insights into the effectiveness of the proposed intrusion detection system, emphasizing its potential in identifying and mitigating security threats in network traffic.