A Hybrid CNN–LSTM Framework for Network Intrusion Detection with SMOTE Balancing

Abstract

In order to improve intrusion detection in network traffic analysis, this research presents a novel hybrid deep learning approach for feature extraction and classification· The complexity, volume, and sequential nature of contemporary network data frequently pose challenges for traditional machine learning techniques, which has a negative impact on anomaly detection's speed and effectiveness· In order to tackle this issue, our approach utilizes a hybrid model that initially automatically extracts significant temporal patterns and spatial features from unprocessed network traffic data using a Convolutional Neural Network (CNN)· Following that, a Long Short-Term Memory (LSTM) network receives these learnt properties and uses its capacity to process sequences to accurately classify traffic and identify interference· Our model outperforms traditional machine learning classifiers like Support Vector Machines and Random Forests by a considerable margin, achieving an accuracy of 98·7% and an F1-score of 97·4% on the publicly accessible CICIDS2023 dataset· The findings show that our hybrid CNN-LSTM model offers a promising, data-driven solution to a significant network security concern by having the ability to greatly improve the speed and efficacy of current intrusion detection systems.