An AI-Enabled Deep Learning Framework for Real-Time Network Intrusion Detection and Cybersecurity Threat Analysis

Abstract

Cyber threats are more dangerous than ever and are growing at a level faster than we can keep up with due to the digitalized world we now “find ourselves in. Most contemporary security systems fall short in this ever-changing environment. Most intrusion detection systems (IDS) rely on past events to establish rules and rely on attack signatures. This means they typically cannot recognize unknown attack patterns or adapt to new techniques when systems become compromised. For this reason, this paper proposes an artificial intelligent (AI) system using deep learning techniques to create a real-time IDS with the ability to analyze the current state of cyber threats. Our proposed system analyzes the flow, the packets, and their behavior to identify attacks and the nature of cyber threats. To assess the cyber threat nature and to identify attacks, we have adopted a comparative modeling processes by utilizing the popular machine learning (ML) algorithm frameworks such as the Supervised Random Forest, the Supervised Logistic Regression, and the Support Vector Machine (SVM), in addition to an Artificial Neural Networks (ANN) system. Preprocessing datasets via normalization and encoding were implemented to yield a high performance and consistency. This completed the framework for the deep learning models of our IDS. Multiple hidden layers with regularization techniques to assist reduce overfitting and improve generalization were built in. The model performance was evaluated using standard metrics and benchmarked against competing systems. The proposed system was constructed as a web-based application with the ability to provide real-time prediction and offer an improved interface to the end-users. We have constructed an intelligent and adaptable system that poses a scalable cyber threat in an ever-digitalized world.