Securing Academic Library VPN Access with AI-Powered Anomaly Detection Models

This study explores AI-powered anomaly detection to secure academic digital library access via Virtual Private Networks (VPNs). A three-model framework One-Class SVM, Isolation Forest, and Long Short-Term Memory (LSTM) Autoencoder analyzes encrypted VPN traffic using the Information Security Centre of Excellence (ISCX) -VPN dataset. Key steps include feature extraction, anomaly detection, and evaluation using Precision, Recall, and F1-Score. Results show the LSTM Autoencoder outperforms others with an F1-score of 0.61, effectively identifying anomalies in encrypted data. The study presents a novel approach to cybersecurity in academic environments, offering a scalable and intelligent solution superior to traditional systems that often fail to detect threats within encrypted VPN traffic.