Machine Learning Based Network Issues Classifier, Recommender and Threat Prediction for OSI Model Layers

Abstract

Computer networking creates remarkable impacts in solving modern activities, but as they have evolved to become larger and more complex, they also present challenges when it comes to finding problems, identifying issues, and examining potential security problems. Most network management systems offer a single focus point, or a technical solution that is not easy for ordinary person to see, leading to time delays, increased risk, and additional vulnerabilities. The goal of this study is to allow for better identification of problems, propose a set of applicable solutions, and find possible threats that may arise through analyzing traffic on the network using an intelligent, machine-learning based model for classifying network problems based on the OSI model. To identify and classify problems, two sets of data were used; real-world database of network problems received from the Office of Accountant General for the Federation (OAGF), and the well-known NSL-KDD datasets for conducting research on intrusion detection. Machine learning algorithms have been examined for each component of the research, with the Naïve Bayes algorithm giving the highest classification accuracy, while Adaboost gave the most accurate classification of malicious traffic. The three models were integrated into a single web application using the Python Flask, HTML and CSS languages, and were designed with a three-layer architecture which splits the interface, processing, and storage components. The full System offers users information about the problem they are having. It also gives them information about the layer on the OSI model that contains the problem as well as recommendations on how to fix it. The system checks if the specific traffic has evidence of an attack. In this paper, showed how machine learning can assist with network troubleshooting, increase security awareness among users, decrease dependence on subject matter experts, and ultimately accelerate and increase accuracy in decision making.