Deep Inductive Transfer Learning Approach for Network Attacks Detection

In recent years, we have been witness to a remarkable surge in technological advancements, accompanied by an unprecedented influx of network data within mobile and IoT applications. Consequently, the exponential rise in network connections has given rise to a concealed threat of potential network breaches. In order to address this danger, it is imperative that we actively strive to devise more effective techniques for monitoring network traffic. A significant challenge in identifying network attacks is their constantly evolving nature, necessitating an adaptable learning approach that can keep pace with these rapid changes. The remarkable advancements in AI/ML research provide a commendable and comprehensive approach to address the challenge of network intrusion detection. In this article, we propose a Transfer Learning model and undertake experiments on simulated network dataset to demonstrate the model's capacity to detect anomalous packets while successfully retaining old knowledge. To construct dataset, we build a VLAN to meticulously simulate various network attacks and meticulously analyze simulated attack packets using an IDS tool named Suricata.