Abstract

In the dynamic landscape of network security, the development of intelligent intrusion detection systems relies on the availability of robust datasets that faithfully emulate real-world scenarios. This study proposes a modified residual connection enabled convolutional neural network framework (MRCECNN) tailored for network intrusion detection, aiming to enhance classification accuracy and robustness in identifying potential threats. By leveraging state-of-the-art techniques in deep learning (DL) and introducing residual connections along convolutional blocks, MRCECNN aims to significantly enhance classification accuracy and robustness in identifying potential threats within network traffic. Its architecture is meticulously designed to analyze intricate patterns in network data by incorporating skip-connection-based residual networks to mitigate the problem of vanishing gradients, enabling it to distinguish between normal network behavior and malicious activities with high precision. The proposed model demonstrated metric values of 0.996, 0.998, 0.997, and 0.996 for accuracy, precision, recall, and F1 score, respectively.

Authors

Shravan Venkatraman; Brindha V; Abeshek A; Aravintakshan S A; Anirudh Vinodh