ANALYSIS OF ENCRYPTED VPN TRAFFIC USING TENSORFLOW

This project report is organized into five chapters, each focusing on the study and implementation of VPN Traffic Analysis using TensorFlow. Chapter One introduces the study by presenting the background, problem statement, objectives, scope, and justification of the project, emphasizing the role of anomaly detection in encrypted VPN traffic. Chapter Two reviews related literature and previous works on VPNs, encrypted traffic analysis, anomaly detection, and the application of deep learning frameworks such as TensorFlow for cybersecurity. Chapter Three discusses the methodology adopted for the project, including Wireshark data capture, preprocessing, feature engineering, and the design of a TensorFlow based neural network for anomaly detection. Chapter Four presents the practical implementation, showing how VPN traffic was captured, transformed into structured datasets, analyzed using TensorFlow, and evaluated for accuracy. Graphs, charts, and model performance metrics are provided to illustrate results. Chapter Five summarizes the findings, draws conclusions, and provides recommendations for future improvements, including the integration of advanced models such as LSTM networks, autoencoders, and hybrid approaches for improved VPN anomaly detection. This project demonstrates that even though VPN traffic is encrypted, valuable insights such as packet size, timing patterns, and protocol usage can be analyzed with the aid of TensorFlow to detect anomalies and potential vulnerabilities. The results highlight the potential of deep learning as a powerful tool for enhancing VPN security and monitoring in realworld applications.