Wifi Goblin - Intrusion Detection Website

Abstract

The intrusion detection system that utilizes Convolutional Neural Networks (CNN) and ANOVA feature selection to identify and classify network activities. The proposed system processes data to detect five types of network behavior: Probe, U2R, R2L, DoS, and Normal. By employing ANOVA for feature selection, the model optimizes the input dataset to 27 critical attributes, enhancing its predictive capabilities. The CNN model achieves an outstanding accuracy of 97%, significantly outperforming Deep Neural Networks (DNN) at 95.11% and Decision Tree classifiers at 94.45%. The system is implemented as a Python Django-based web application that allows users to upload CSV files, facilitating real-time detection and classification of intrusions. This approach combines state-of-the-art machine learning with an intuitive interface, addressing the critical need for robust and user-friendly network security tools. Additionally, the paper highlights the impact of ANOVA feature selection on model performance, the superiority of CNN in handling complex datasets, and the practical implications of this solution in real-world cybersecurity applications.