The AI-Based Internet Routing Anomaly Predictor is an advanced machine learning-driven system designed to detect abnormal patterns in network traffic and identify potential cyber-attacks in routing environments. With the rapid growth of internet usage and increasing complexity of network infrastructures, traditional security mechanisms such as firewalls and signature-based intrusion detection systems are often insufficient to detect unknown or evolving threats. To address these limitations, the proposed system utilizes the NSL-KDD dataset, a widely recognized benchmark dataset for intrusion detection, to train and evaluate machine learning models. The system incorporates data preprocessing techniques including data cleaning, normalization, and feature encoding to convert raw network traffic data into a structured and machine-readable format. Classification algorithms such as Random Forest and Decision Tree are employed to analyze traffic patterns and classify them into normal or anomalous categories with high accuracy. Among these, the Random Forest algorithm demonstrates superior performance due to its ensemble learning approach, reduced overfitting, and improved generalization capability. Furthermore, the system is integrated with a Flask-based backend and a user-friendly web interface developed using HTML, CSS, and JavaScript. This integration enables real-time monitoring, log file uploads, anomaly detection, and alert generation. The system also maintains a SQLite database to store user data, network logs, and historical anomaly records for further analysis. The proposed solution significantly enhances network security by automating the detection of suspicious activities, reducing human intervention, and improving response time. It provides a scalable and efficient framework suitable for enterprise networks, cloud infrastructures, and educational institutions. The system demonstrates the effectiveness of machine learning techniques in improving cybersecurity and offers a foundation for future enhancements such as deep learning integration and real-time streaming analysis.
Talwar et al. (Thu,) studied this question.