AI-Powered Cyber Intrusion Detection System with Real-Time Threat Analysis and Web Application Interface for Enhanced Network Security

This research paper presents a novel idea for an AI-powered Cyber-intrusion detection system that combines the architecture TabNet with Deep Neural Networks (DNNs) to enhance network security. The Tabnet utilises a sequential attention mechanism for dynamic feature extraction and selection, and the selected features are fed into a deep neural network for accurate prediction. The Tabnet prioritizes important features at every decision point, providing Interpretability through an attention mechanism that improves it by highlighting the most important features that influence predictions. The extracted feature subset is then fed to a deep neural network, which is well-suited for learning and identifying complex patterns in an optimized feature space. The proposed approach works well with high-dimensional network traffic data, improving detection performance. In this proposed approach, the system is also deployed with a frontend web application using Python’s Flask library to make predictions using a customised dataset. Users can enter the values of specified features to predict the type of cyber-attack. The model is evaluated on the CICIDS 2018 dataset, which consists of 79 features and 24,67,820 network data, and has also obtained an accuracy of 94.22%. The proposed system outperforms traditional machine learning models in intrusion detection.