CNN-RBM Integrated Deep Learning Design for Categorizing Attack in an Intrusion Detection System
A. Kalaivani, R. Pugazendi · International Journal of Uncertainty Fuzziness and Knowledge-Based Systems · 2025
Currently, network attacks and intrusions are increasing due to expansions in computer networks. The most critical issue these days in modern cyber networks are network attacks. Intrusion prevention systems are designed to enhance security along with the firewalls and other intrusion prevention systems. Each and every network regardless of its size is exposed to network attacks. An Intrusion Detection System (IDS) is an essential security tool for categorizing malicious attacks in networks. Presently, Machine Learning (ML) and Deep Learning (DL) models are applied for developing a competent IDS and in numerous domains. Automated detection of malicious attacks in a timely manner is the purpose of IDS. Advanced cyber security solutions are required for continuous detection of malicious threats. Hence, investigators are generating an effective IDS for this research problem due to complex malicious attacks. In this article, an integrated DL model comprising of Convolutional Neural Network (CNN) with Restricted Boltzmann Machine (RBM) are applied to generate a fusion IDS to predict and classify malicious attacks. In the proposed Integrated Convolutional Restricted Boltzmann Machine Intrusion Detection System (ICRBM_IDS), the CNN executes convolution to hold local features and RBM captures the temporal features to enhance the performance of Intrusion Detection and Prediction. The ability of the ICRBM_IDS model is assessed based on the ID data present widely. The experiments were conducted on CSE-CIC-DS2018 dataset which is the result of collaborative project between Communications security Establishment (CSE) and the Canadian Institute of Cybersecurity (CIC) that is currently used and realistic. The simulation results of the proposed ICRBM_IDS outperform the present ID methods by attaining a high accuracy rate by detecting malicious attacks.