An Integration of Proxy Servers with Machine Learning for Imporving Security to Networks in Campus Wireless Network Systems

Since educational institutions and organizations have relied on campus wide Wi-Fi, secure internet access has become more important with time. In this paper, we describe a proxy server-based system to monitor and control web access while using machine learning (ML) to predict potential cyber threats. Traffic is regulated by the system using a proxy server on a gateway machine, Squid. If a user asks to access a website, the proxy sees if the domain is on the blacklist. If it is, access is denied. If not, the request is then forwarded to an ML model trained on the UNSW-NB15 dataset to determine if there is a risk. The attack types in this dataset range from DoS, Worms, and Reconnaissance to give a complete coverage of threat detection. Machine learning component is comprised of Random Forest and XGBoost classifiers that have high recall rates to minimize the threats missed. With an admin dashboard, logs and traffic details are available for real time monitoring and decision making. The access control and predictive analytics are combined in this system to provide a scalable and reliable solution to network security. It integrates traditional proxy methods with cutting edge ML models for safe internet usage and protection users from evolving cyber threats in dynamically evolving network environments.