A Hybrid Approach of Data Mining and Deep Learning for Network Intrusion Detection

The growing complexity and volume of network traffic poses significant challenges to traditional intrusion detection systems (IDS), often leading to inefficiencies in detecting unauthorized access and malicious activities. To identify different types of network attacks, many IDS have been proposed using artificial intelligence or machine learning, but the results are still not satisfactory for most of these systems. To improve intrusion detection in large-scale networks, this research study proposes a hybrid approach combining data mining and deep learning. Feature selection is performed using Minimum Redundancy Maximum Relevance (mRMR) to identify highly relevant features to the target variable and have the minimal redundancy of those selected features calculating the mutual information. The Convolutional Neural Network (CNN) with 1D convolutional layers detects small, localized complex patterns by adopting structured data. This approach efficiently handles large volumes of data while maintaining high accuracy. This system achieved an impressive detection accuracy of 99.42%, showcasing the potential to combine advanced techniques for the detection of network intrusions.